Water & Wastewater Treatment

www.wwtonline.co.uk | WWT | MARCH 2020 | 11 the output criteria. It means they can recommend the best course of action in any situation, even in the face of rapidly changing circumstances." Q. What kind of decisions can they support? JG: "Xylem's neural networks, which are currently in operation in between 30 and 40 German plants, optimise energy consumption and chemical usage, while increasing compliance margins. "They use real world data collected for all salient parameters involved in processes – such as carbon, nitrogen and phosphorus elimination – to build models. "These models, or "digital twins", can run thousands of simulations which are then used to predict energy and resource use in any given situation. Plants can use this information to set targets for opera- tion: to favour the lowest energy level possible, or the highest elimination rate, for example. "It means that neural networks can support any number of real-time decisions. That might be, for example, heightening the efficiency of aeration to cut oxygen wastage, or optimising energy consumption to make savings." Q. How are neural networks dif- ferent to operator-led optimising processes? JG: "Neural networks use the same strategies and ask the same questions as human operators when they are making decisions. Success-limiting parameters, critical effluent parameters, expected loadings and boundary conditions, for instance, will all be factored in by the artificial intelligence. "A neural network will analyse pat- terns 24 hours a day, seven days a week, and base its decisions on predictive math- ematical models. "This allows plants to move away from wasteful, reactive management and in- stead work towards proactively optimising processes as conditions change." Q: How does it work in practice? JG: "At Germany's Cuxhaven Wastewater Treatment Plant, managers wanted to re- duce energy consumption for aeration in five parallel biological tanks, while com- plying with legal effluent concentrations. To help them achieve this, Xylem's BluX neural network team used the plant's real- world data to build a model of the carbon, nitrogen and phosphorus elimination processes. The digital twin was used to determine the best setpoints for the aera- tors in each zone, and the team used vir- tual analysers to calculate the incoming carbon, nitrogen and phosphorus loads. The resulting optimisation strategies led to a 26.3 per cent, or €330,000, reduction in energy use over the course of a year." Both Ruth Clarke and Jöerg Gebhardt recently presented a webinar entitled 'Neural networks: teaching your treatment works to be smarter'. For more information visit www.xylem.com/uk Participants: Ruth Clarke is business development manager in advanced infrastructure analytics at Xylem Water Solutions UK. Dr Jöerg Gebhardt is managing director of Xylem brand Acquatune. PREVENT LEAKAGES; SAVE WATER! TEL: +44 (0) 208 670 7511 FAX: +44 (0) 208 761 2456 EMAIL: mail@denso.net www.denso.net WINN & COALES (DENSO) LTD A MEMBER OF WINN & COALES INTERNATIONAL INCREASE THE SERVICE LIFE OF YOUR WATER PIPELINES AND FITTINGS WITH DENSO ® • OUR CORROSION PREVENTION SYSTEMS HELP TO REDUCE THE RISK OF COSTLY REPAIRS AND WATER LOSS • FULLY COMPLIANT WITH P1 & P2 CIVIL ENGINEERING SPECIFICATIONS FOR THE WATER INDUSTRY • FREE APPLICATION TRAINING AVAILABLE • CONTACT US FOR MORE INFORMATION BS EN ISO 9001:2015 & 14001:2015

• Cover
• Contents
• The Talk
• In Focus
• The Knowledge