Boundary forecast

The historical demand varies with a clear daily cycle and a less pronounced weekly cycle.

The demand forecast is an important input to the optimisation.

It is one of the project's visions to experiment with different types of models to predict the forthcoming demand.

AI developed three machine learning models, and these are compared to a simple baseline developed by AaV.

Zone exchange forecast

In addition to the demand, each pressure zone in a water distribution network may also be connected to neighbour zones. The exchange with neighbour zones is typically decided on a day-to-day basis, and is less predictable than the cyclic demand.