This demonstrator will focus on a network of ultrasound water-level measurement sensors used for the early flood warning in the Athens metropolitan area. During the past few years, the Athens metropolitan area has been affected by sudden and very heavy rainfall that renders systematic prevention effortless, due the transitional behaviour of the event. Nevertheless, the water and sewage company (EYDAP SA) utilises a network of sensors to better manage these types of events and rapidly deploy anti-measures consisting of tankers rushing towards the flooding sites trying to resolve the problem.

FUELICS ultrasound liquid level monitoring sensors have already been deployed (less than a dozen) within the sewage network of the city of Athens and are installed under the manhole lids in order to measure the level of underlying rain or sewage water. The city’s sewage network is many kilometres in length. The sensors are mostly deployed in locations where a high frequency of flooding has been observed, and a few are deployed upstream of these locations to provide additional measurements and early warnings of possible flood events.

In the case of a heavy rainfall, the time window the authorities have between detecting a possible flood event and reacting is currently in the order of 10 minutes, which is the time it takes them to deploy on-site anti-flood measures. For this reason the FUELICS sensors are currently configured to connect and transmit data every 5 minutes to the authorities in order to inform them as soon as possible of expected floods. This configuration leads to a waste of energy on the sensors because they send data very frequently even when there is zero probability of flooding, which dramatically decreases the battery life. Furthermore, the fact that each sensor operates autonomously, leaves all the data that are collected from the network of the upstream deployed sensors unutilised.

Added value by Swarmchestrate: The goal of the demonstrator is twofold. Firstly, to deploy an application that will dynamically reconfigure a large network of sensors, around 100 of them, using data from external sources like weather forecast and weather statistics databases to optimise the measurement and connection interval of each sensor, and maximise the battery life without decreasing the capability to detect a flood event. The application will also use the external data to dynamically create Swarms of these application components (deployed as containerised microservices) in the Cloud-to-Edge continuum that will combine the measurements and use the upstream data to forecast a possible flood event before it actually occurs and gets detected by the physical sensor, thus providing to EYDAP a longer reaction window. The application will be able to use the sensor-containers to send reconfiguration commands to the sensors and physically change their behaviour as needed.

Secondly, we will create large scale simulations of deployments, including hundreds to thousands of sensors in the Athens metropolitan area, using historical data of rainfall from previous years. The aim is to validate the operation of the application in an even larger scale and assess the effectiveness of the deployments in a simulated environment which will allow authorities to have more confidence in selecting the proper location for the installation of the sensors before the actual rollout. This implementation will be utilising the decentralised simulation environment developed by Swarmchestrate. A digitally signed support letter from EYDAP is provided.