Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


14023.pdf (2.9 MB)

ETD Abstract Container

Abstract Header

Stochastic Demand-hydraulic Model of Water Distribution Systems

Chen, Jinduan
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439301579

Abstract Details

2015, PhD, University of Cincinnati, Engineering and Applied Science: Environmental Engineering.
Recent years, the expansion of Supervisory Control And Data Acquisition (SCADA) systems and the installation of Advanced Metering Infrastructure (AMI) have become common in water utilities, thanks to the advances in telemetry technologies. As a result, research interests have been focused on how to utilized these online data sources to improve system operations. However, as an important means of predictive analysis, the traditional hydraulic models could not be applied in a real-time scenario due to the unobserved, spatially distributed water demands. This study presents the research efforts on an stochastic demand-hydraulic model that assimilates hydraulic observations and estimates water demands in real-time. The water demands are modeled as a statistical time series with the time steps aligned with the Extended Period Simulation (EPS) of network hydraulics. In the first part of the research, the Box-Jenkins methodology was applied to short-term water demand series. Two observed data sets of aggregated water demands were collected and analyzed. The double-seasonal auto-regressive models were identified for both data sets. The time series models were estimated and utilized for the forecasting of short-term water demands. Results show that the time series models produce more accurate forecasts than the traditional models of demand patterns. A software toolkit for the multi-seasonal time series models was developed in the MATLAB programming language. The software implemented the methods for the identification, estimation, forecasting, and diagnostic checking. The adaptive parameter re-estimation algorithm of Generalized Likelihoods Ratio (GLR) was also developed and tested. Results showed that GLR improves the accuracy of the probability limits of the forecasts. In the second part of the research, a composite demand-hydraulic model was established by linking a time series water demand sub-model to the network hydraulic sub-model. The Expectation-Maximization (EM) algorithm was suggested to estimate the model parameters and water demands in an iterative manner. The EM algorithm was applied to a network with seven customers with unknown water demands. The convergence in parameters was achieved in fifteen E-M iterations. Analysis of the results showed that the estimated water demands have the absolute relative errors of 7%-10% on average. Both temporal and spatial correlations of water demands can be obtained by studying the demand estimates. A software toolkit featuring the demand-hydraulic model was developed in C/C++ for the real-time modeling of distribution networks. The toolkit included six modules for database access, time series modeling, hydraulic modeling, SCADA generation, EM-based estimation, and visualization. Overall, the study showed the applicability of the statistical time series in the modeling of both total water demands and spatially distributed water demands. The method and software developed in this study can utilize the incomplete and indirect hydraulic measurements to produce demand estimates in a real-time context. As a result, the SCADA systems and the hydraulic models were integrated under the proposed framework that is capable of supporting predictive analysis. The proposed stochastic demand-hydraulic model is the foundation for many future research directions such as operational optimization, data collection system design, consumer node clustering/grouping, and development of real-time applications.
Dominic Boccelli, Ph.D. (Committee Chair)
Michael Eugene Tryby, Ph.D. (Committee Member)
James Uber, Ph.D. (Committee Member)
Lilit Yeghiazarian, Ph.D. (Committee Member)
151 p.
Environmental Engineering
Water distribution system; Real-time hydraulic model; Water demand; Time series forecasting; Parameter estimation

Recommended Citations

Citations

  • Chen, J. (2015). Stochastic Demand-hydraulic Model of Water Distribution Systems [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439301579

    APA Style (7th edition)

  • Chen, Jinduan. Stochastic Demand-hydraulic Model of Water Distribution Systems. 2015. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439301579.

    MLA Style (8th edition)

  • Chen, Jinduan. "Stochastic Demand-hydraulic Model of Water Distribution Systems." Doctoral dissertation, University of Cincinnati, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1439301579

    Chicago Manual of Style (17th edition)

ucin1439301579
379
© 2015, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.