A post by StormHarvester
StormHarvester is now coming to the end of our three-month intelligent sewer challenge currently being run by Wessex Water in the Bath catchment. As part of this challenge we are monitoring all combined sewer overflows (CSOs) and pumping stations in the Bath catchment in real time and alerting operational staff to potential blockages prior to that blockage turning into an unpermitted out-of-sewer (overflow or spillage) event. As part of this challenge we are receiving data from Wessex Water’s existing CSO and pumping station level sensors and running continuous analytics on their levels to determine in real time when and where a sewer blockage may be forming.
Using our machine learning prediction system, we are predicting a safe operating window for each sewer over the upcoming six hours. Our machine learning system is linked to a meteorological office six-hour hyperlocal radar rainfall prediction system in order to accurately adjust the expected level of each CSO and pumping station when rain is on the horizon.
Through the course of the trial we have been sending regular automated alerts directly to Wessex Water when potential blockages or level anomalies are identified. The graph below shows an example of a blockage alert that was sent in mid-July during the trial. From the graph you can see that the StormHarvester system sent an alert on a potential blockage several days before the sewer reached the overflow level.
In general, we have found our interaction with Wessex Water staff through the Marketplace challenge to be very positive. With the assistance of their technical staff, the StormHarvester system was set up and made operational within days. A designated member of the Wessex Water operational team has been available throughout the trial to provide feedback on our anomaly and blockage alerts, occasionally even providing us with photos of the debris that caused the blockages after we had alerted Wessex Water and they had successfully removed the blockage. The open nature of our dialogue with the relevant Wessex Water staff has allowed us to refine our machine learning models quickly during the trial, so only a handful of genuine alerts are now being sent to Wessex Water each day including during periods of heavy rainfall.
StormHarvester’s sewer monitoring system is designed to give sewer managers peace of mind knowing their sewers are being monitored 24/7. The system is designed to reduce the overall number of non-permitted spillage events by identifying early-stage blockages or unexpected high sewer levels and allowing operational staff to be deployed prior to an unpermitted spillage or flooding events occurring.