AMIS helps a water authority reduce its CO₂ footprint through machine learning
A machine learning model that combines water levels, electricity prices and weather forecasts to optimise pumping operations through predictive planning.
Optimising pumping station operations based on weather forecasts and electricity prices
A water authority is responsible for water quality and safety in a particular region. It aims to keep water levels as stable as possible, supplying additional water during dry periods and ensuring that sufficient water is drained away in time during wet periods.
Water supply and drainage are managed by pumping stations with a relatively simple control loop. When water levels are too low, the station pumps water into the area; when levels are too high, it pumps water out. Pumping consumes a lot of electricity, and responding to changes in water levels is a reactive process.
Predictive pumping
The water authority asked AMIS to develop a machine learning model that optimises pumping based on water levels, electricity prices and weather forecasts. This enables predictive pumping, for example when heavy rainfall is expected later in the day. A pumping station can also be activated proactively when electricity prices are favourable.
Sustainability and reliability
Validation of the model shows that strategically activating a pumping station earlier or later can help maintain more stable water levels. Another benefit is that pumping stations can operate when there is a surplus of renewable energy, potentially reducing the water authority’s CO₂ footprint considerably. Finally, optimising pumping operations around variable electricity prices delivers significant cost savings while maintaining the same quality and safety of water management.
