10.15131/shef.data.5364100.v1 S. M. Masud Rana S. M. Masud Rana Paulo José Oliveira Paulo José Oliveira Tian Qin Tian Qin Dominic L. Boccelli Dominic L. Boccelli CCWI2017: F123 'Case Study: Improvements to a Real-Time Network Modelling Framework' The University of Sheffield 2017 CCWI2017 real-time modeling time series clustering Civil Engineering not elsewhere classified 2017-09-01 15:11:01 Journal contribution https://orda.shef.ac.uk/articles/journal_contribution/CCWI2017_F123_Case_Study_Improvements_to_a_Real-Time_Network_Modelling_Framework_/5364100 Short-term water demand forecasts are valuable for distribution system operators controlling the production, storage and delivery of drinking water. In certain problems, such as real-time pump scheduling, the cycle of data acquisition, model computation, and decision-making is time-sensitive, and requires an automatic procedure to handle the transfer of information between data source(s), forecasting model(s), and the operator. Recent development of a composite demand-hydraulic model integrates a demand time series model with a hydraulic network model to estimate and forecast demands using measurements typically available to water utilities. The application to a real-world network model with approximately 12,000 demand nodes and six flow measurements resulted in good representation of the observed flow rates. However, the performance of the demand-hydraulic algorithm, and subsequent analysis, has demonstrated limitations in two aspects of the demand estimation and forecasting framework: the temporal representation of the estimated demands, and the clustering approach needed to reduce the scale of the parameter estimation problem. The current research will present preliminary results associated with data-driven approaches for representing the temporal demands and application of alternative clustering algorithms to improve the overall demand estimation process.<br>