Sim City: agent-based water demand model


  • Start date
    15 Nov 2019
  • End date
    31 Dec 2020
  • Principal
  • collaborating partners
    Evides, TU Delft

One of the biggest and most significant uncertainties in the development of network blueprints is the future water demand. Nonetheless, this uncertainty receives only limited or insufficiently detailed attention in today’s design process, both in practice and in research. The objective of this project is to study the applicability of agent-based models for the drinking water sector in the modelling of supply areas and the associated future water demand.

Knowledge about agent-based models

Water demand scenarios in the current design process are based on the water demand measurements over the preceding years. What is lacking are well-founded and concrete future scenarios. In order to make such forecasts reliably, we need knowledge about the development of the water demand per person, but also about the development of cities and of areas as a whole – for example, where people will be working and living.

Micro-simulations, or agent-based models, operate at the level at which all sorts of factors that influence decisions can best be assimilated: the behaviour of individuals, such as a person, a family or an enterprise. From the interplay between all individuals – person A, when confronted with specific conditions and options, chooses A, person B chooses B – there emerges a behaviour that applies to the city or the area as a whole: the macro level. Such agent-based models are already being applied in various contexts, such as for traffic or in the modelling of household energy consumption. With this in mind, it is interesting to examine whether the models could also be valuable for the drinking water sector in modelling future water demand.

It is anticipated that agent-based models will provide insight into a variety of factors, which might vary from one region and/or district to another: the direction and extent of changes in the drinking water demand over time, the uncertainty in these values, and whether the biggest change can be expected to occur at the beginning or at the end of the lifespan of a pipe network. An agent-based model can reveal both variations and consistencies between regions and districts simultaneously. Thus, the water demand in region A can increase while that in region B can decrease; or early adopters (e.g., people who install water-saving toilets sooner) live in district C, which is not the case in district D. The foundations for such a demand model have already been established in the Joint Research Programme of KWR and the water utilities over the last ten years: all the basic data are collected in SIMDEUM, making possible the modelling of the water demand of individual agents.

Overview and functional design

First of all, the project involves the conduct of an overview of the existing agent-based models and applications, both in the Netherlands and internationally. This is followed by an inventory of the possible applications within drinking water distribution.

The second step involves producing a functional design in which the following aspects are highlighted:

  • data and data streams required to create an agent-based model;
  • structure for modelling of the demographic, city and district development;
  • establishment of a structure to forecast water demand based on the demographic, city and district development, with the possible option of incorporating the data from the agent-based model into SIMDEUM.

The proposed subject has also opened leads for collaboration with TU Delft. Between November 2019 and February 2020, students will be developing an agent-based model aimed at simulating the future development and associated water demand of the city of Delft. It is expected that the results of the collaboration with TU Delft will nicely illustrate how an agent-based model could work for the drinking water sector.

Possibilities for the drinking water sector

The result of this research project will clarify the possibilities of using agent-based models for the simulation of future water demand scenarios and, through a phased implementation plan, indicate the requirements for building such models. This will open up new possibilities in the design of future-proof and robust network blueprints and scenario studies. It will give water utilities greater insight into the level of future-proofness of their existing distribution networks, and of the already designed network blueprints, and how the latter might need to be modified.