Technology

Wessex Water trials AI simulation tool

20 October 2020

Wessex Water has piloted an artificial intelligence (AI) system for drainage water management plans (DWMPs).

The system, which uses Modefrontier simulation software with EnginSoft, took just four days of an engineer’s time and 12 days’ CPU time to evaluate thousands of solutions.

"We’ve been deploying AI in the automotive sector for a number of years and believe Modefrontier is the most versatile software for smart optimisation," said Bipin Patel, EnginSoft UK’s managing director. "We approached Wessex Water with our Strategicgiant flood assessment software, which is ideal for assessing flood scenarios as part of the growth impact planning process. However, since they were wanting to tackle DWMP generation, direct optimisation with Modefrontier was the most appropriate approach for Wessex Water."

All UK water authorities are required to develop DWMP for all catchments as part of AMP7. Wessex Water took the strategic decision early on to trial suitable software.

Patel said: "Modefrontier’s ability to integrate a cost model for each solution is valuable, which was a deciding factor in Wessex Water’s choice. It cuts out what we call ‘optioneering’: the trial and error generation of plans which are then found to be overly costly.

"Modefrontier substantially reduces the time it takes to assess thousands of solutions that are costed from the outset and provides tools such as the Pareto Front to understand the trade-off between total cost versus quality of solution – in this case, how much flooding you can eliminate."

David Searby, Wessex Water’s wastewater modelling technical manager, said: “The EnginSoft trial for DWMP optioneering delivered positive results and we have now taken the decision to purchase a one-year licence to allow us to look at more catchments using this AI-based approach.”

Once catchment details were captured (covering the likes of storage locations, pipe upgrades, SUDS areas, flood monitoring locations), the Modefrontier workflow was generated in a matter of hours. The optimisation study then ran for four days and outputted a series of graphical results with varying cost-versus-benefit trade-offs.

It was here that EnginSoft was able to provide some surprises for the Wessex Water engineers.

"We could see that there were a number of options on the far side of what we call the ‘Pareto Cliff’, where if one or two constraints were tweaked, a significant reduction in cost could be achieved. It's not uncommon for restrictions to exist undetected in complex systems for years, limiting performance, and I’ve seen them highlighted by AI before. The guys at Wessex were fascinated to see the analysis, and I’m sure there’ll be similar instances highlighted as Modefrontier usage increases," Patel concluded.