The results of the Naiades project will allow users to not only detect water leakages, but to accelerate their detection thus helping water utilities to react faster to potential problems.
The initial results of the Naiades project will be tested with pilot partner CUP Braila, a regional public water and sanitation operator in Romania. The approach developed by Naiades will allow the user (in our case CUP Braila) the access to advanced analytics enabling better monitoring of the water distribution networks in real-time. This will allow the user to react much faster to anomalies in the water distribution system. As part of NAIADES, temporal, spatial analysis and analysis of states have been developed to detect anomalies faster. Let us have a look at how each of the approaches works and how they have been tested in Braila, Romania.
Three Naiades approaches in a nutshell
Temporal analysis is an approach that searches for anomalies based on given time series. The first step is to break an incoming raw signal into a series of pre-processed signals, which are then all pushed into the anomaly detection engine. The detected anomalies are then aggregated into one meta signal which serves as a base signal for distinguishing relevant anomalies from irrelevant anomalies.
We are currently profiling the anomalies based on the time series provided by CUP Braila so that we can separate the relevant anomalies from the irrelevant ones.
This approach can identify approximate and precise leakage locations. The approximate leakage location can be determined by comparing theoretical and actual pressure values, since there is a pressure drop when leakage occurs. The accurate leakage location approach is done by scanning a selected area with mobile noise sensors - and based on the information collected, a leak location is assessed.
Accurate leakage detection is performed by scanning a selected area with 4 mobile noise sensors. Once the noise sensors are installed in the field and the search is triggered, the algorithm relocates the sensors until it has collected all the necessary data. As a result, you get an estimate of where the leak might be. We have started spatial testing of (fully automated) algorithms to accurately locate leaks in the water network in Braila.
State analysis can enable the detection of an anomalous event and the transition rate to the anomalous state. In addition, users can see the probability of transition from a pre-normal state to an abnormal state.
This approach also allows users to predict potential anomalies before they even occur! Since we can assign a trigger to each state, we can also assign a trigger to the pre-anomalous state. So, when a real-time sample enters the pre-anomalous state, a trigger is triggered to warn the end-user of the possible transition to the anomalous state.
Right now, all three approaches are being tested! Stay tuned for all the results!