Water quality, aging infrastructure, water scarcity in arid areas as well as increased demand raise the need for proactive water-related risk management. By bringing together the strengths of sensor networks and a state-of-the-art, distributed, programmable middleware, WP7 provides a holistic approach to the real-time management of complex water systems. By applying a distributed infrastructure of sensors and actuators we achieve, continual monitoring of the water network and its components, e.g., pipes, catchments, valves etc. The readily assimilated information which is produced by monitoring and that may be related to flooding, drought, water quality or infrastructure outages will be used in order to plan corrective action and real-time reconfiguration of the network, whenever needed. For example, if it is determined that rainwater contains a very low amount of TDS (Total Dissolved Solids) causing, among other things an unpleasant odor, it may be desirable to mix this water with water from other sources with higher TDS. By programming such a reactive policy in the sensor network’s middleware, as soon as the previous situation is detected by a water quality sensor, the valve that controls the mixing of surface water and rainwater may be opened automatically with the help of an actuator. As a result, a higher concentration of surface water is mixed in the rainwater, thus increasing the amount of TDS is the mixture. The above modus operandum is equivalent to a closed-loop control circuit. A complementary behavior to the monitor-react cycle described above involves the mining of the monitoring data for the detection of long-term trends or emerging patterns. For example, by applying the naïve Bayes classifier, we can predict, based on historic data, demand, or water quality related values.
As mentioned above, the monitoring of the distributed water network is undertaken by the sensor network, consisting of several sensor nodes. Sensor nodes are deployed in key locations in the network close to the components/devices that need to be monitored, e.g., valves, catchments, pipes etc. Each sensor node employs a set of sophisticated water-related sensors for measuring the value of parameters of interest. Sensor nodes may be diverse in sensing functionality, each employing different sensor types, e.g., water quality, Ph, flow, pressure etc. For example, a pressure sensor reporting very low values may be an indication of a leak. The data that is produced by the sensors is communicated by means of a shared bus to the central processing unit (CPU) of a microcontroller that is embedded in the node. The microcontroller runs the software policies that apply to the received data triggering the appropriate actions wherever needed. Data may also be transmitted for further, computationally intensive processing, to a master node outside the sensor network by means of a low-power, embedded radio transceiver. The above components are powered by a battery that is situated on the node.
Each node samples the water continually, measuring variables that are related to the types of sensors it contains. The measured values may be compared to a predefined threshold, compared against other values (sensor fusion) or against their historic dataseries (trend analysis). Average values, e.g., average demand, average pressure in a segment, are also of interest in this domain as they can be used in order to evaluate the network and predict its behavior under different circumstances of usage.