Keywords
demand-forecasting; R; EPANET; water-distribution-systems
Start Date
25-6-2018 2:00 PM
End Date
25-6-2018 3:20 PM
Abstract
Internet of things sensors for water demand contribute to the big data water modelers can use to characterize and forecast demands at increasingly finer scales in space and time. Popular modeling techniques include artificial neural networks and autoregressive models but other methods such as deep learning and Gaussian mixture models are also in use. Leveraging these demand models to optimize planning and operational decisions often requires incorporating demands into simulations of system behavior. This paper describes solutions for water network simulations and demand modeling in the R environment. A key component is the recently released epanet2toolkit for water network simulations. Examples are provided for using the package for three tasks related to demand modeling. First we update a network with forecasted demands. Next, we connect the updated model to an optimizer in order to minimize energy use. Finally, we carry out Monte Carlo simulations of system behavior driven by uncertainty in demand forecasts. The fact that these examples require only modest code writing should motivate researchers and practitioners to incorporate more demand data into operational and planning decisions.
Water demand and network modeling with R
Internet of things sensors for water demand contribute to the big data water modelers can use to characterize and forecast demands at increasingly finer scales in space and time. Popular modeling techniques include artificial neural networks and autoregressive models but other methods such as deep learning and Gaussian mixture models are also in use. Leveraging these demand models to optimize planning and operational decisions often requires incorporating demands into simulations of system behavior. This paper describes solutions for water network simulations and demand modeling in the R environment. A key component is the recently released epanet2toolkit for water network simulations. Examples are provided for using the package for three tasks related to demand modeling. First we update a network with forecasted demands. Next, we connect the updated model to an optimizer in order to minimize energy use. Finally, we carry out Monte Carlo simulations of system behavior driven by uncertainty in demand forecasts. The fact that these examples require only modest code writing should motivate researchers and practitioners to incorporate more demand data into operational and planning decisions.
Stream and Session
B1: Modelling and Managing Urban Water and Energy Demands in the Era of Big Data