EARLY WARNING OF ATYPICAL SEWER SITUATIONS
How can the Water Vallei and Eem Platform receive signals at an early stage about atypical situations in the sewer system?
Water Valley and Eem Platform is a partnership between the Vallei and Veluwe Water Board and various municipalities in the region. The platform’s software monitors all water flow across municipal sewers and the water board, and provides central control. Data from sensors throughout the network is assessed manually by operators – yet this presents a significant level of work and can mean that certain situations can be overlooked. Such examples include inflowing surface water, a broken sensor or a pipeline blockage.
The client wanted to explore the potential to generate an automatic signal if network values differ significantly from the norm – during heavy precipitation or drought, for instance.
Platform Water Vallei en Eem also wanted to learn more about data science and its various techniques. Together, therefore we implemented Y-Flex, a method to quickly increase data science capacity. Working in partnership with the client’s data analysts, we used a number of techniques to extract more value from the data. In this phase we kept the scope limited and focused on a single area: overflow.
A tool was developed to identify data anomalies. While this was specifically created to detect overflows, with some fine-tuning the tool can also be applied to other water events.
The model predicts water levels in the event of an overflow, based on various trends such as current weather and season. This data is compared with live measurements. If the two values differ substantially from each other, we call this an anomaly. It can signify several things: the sensor may need replacing or something is wrong, such as inflow of surface water or a blockage in the system.
An example of an anomaly: the model expects the water level to drop after the rain shower (blue line), yet the measured water level remains at a higher level (black line). Cause: a blockage.
“The model enables us to recognise sensor failure at an early stage”
The tool means that water situations can be predicted far in advance. Working with the Platform and its members, we are investigating how we can further develop this in new software systems to avoid unnecessary costs and gain even more visibility of the sewer system.