Classification of Water Distribution Systems for Research Applications Steven Hoagland 1, Stacey Schal 2, Lindell Ormsbee 3, Sebastian Bryson 4

Transcription

1 Classification of Water Distribution Systems for Research Applications Steven Hoagland 1, Stacey Schal 2, Lindell Ormsbee 3, Sebastian Bryson 4 1 Graduate Student, Department of Civil Engineering, University of Kentucky, 161 Raymond Building, Lexington, KY 40506, steven.hoagland2@gmail.com 2 Research Assistant, Department of Civil Engineering, University of Kentucky, 161 Raymond Building, Lexington, KY 40506, stacey.schal@gmail.com 3 Director, Kentucky Water Resources Research Institute, University of Kentucky, 233 Mining & Minerals Resource Building, Lexington, KY 40506, ph: (859) , lormsbee@engr.uky.edu 4 Associate Professor, Department of Civil Engineering, University of Kentucky, 161 Raymond Building, Lexington, KY 40506, ph: (859) , sebastian.bryson@uky.edu ABSTRACT Water distribution system models can aid utilities in achieving more reliable and optimal operations of their system. They are also useful in research efforts aimed at improving the planning, design, and operation of systems. This paper outlines the development, classification process, and analysis of 15 water distribution systems for the purpose of creating a database of system models which can be used among the research community to test newly developed algorithms. Differences in basic system characteristics based on configuration are also examined to determine if certain characteristics (e.g. number of tanks, average pipe diameter, etc.) vary systematically by configuration. The study aims to help quantify differences in the three main system configurations beyond the general layout differences. Such a classification may be useful in generalizing the economic performance, reliability, resiliency, or required characteristics (e.g. number of pumps, tanks, etc. per total system demand) of such systems. Such statistics may also be useful in helping to forecast system expansion needs (pipe, tanks, etc.), and security needs (i.e. number of water quality sensors, etc.) as the system continues to grow and expand. 1.0 INRODUCTION Water distribution systems are responsible for providing a clean and reliable source of potable water to communities. System models are typically developed by utilities to perform long-term planning, design new components, and resolve hydraulic or water quality problems (AWWA, 2012). A simple analysis of pressures and flows in a network over time can be helpful in providing information to a utility about the behavior and characteristics of their distribution system. Researchers also utilize models in developing new methodologies and algorithms to aid in planning, design, and operation of systems. This research can range from achieving reliability and optimal operation of a system to water security issues like ideal placement of water quality sensors.

2 This paper discusses the model development, classification process, and analysis of 15 water distribution systems for the purpose of creating a database of system models which can be used among the research community to test newly developed algorithms. The water distribution system models were created in a well-known hydraulic modeling software (i.e. KYPIPE) from standard GIS datasets using a process developed by Jolly et al (2013). Each system was classified as a grid, loop or branch system using new methodology presented herein. A statistical analysis of the database was performed to determine if certain characteristics (e.g. number of tanks, average pipe diameter, etc.) vary systematically by configuration. 2.0 DEVELOPMENT OF WATER DISTRIBUTION SYSTEM MODELS A database of 15 network models was created using a procedure introduced in a previous study (Jolly et. al., 2013) which involves multiple data transfers from various software packages (i.e. ArcGIS and KYPIPE). A general schematic of the model creation procedure is shown below in Figure 2.1. System data files for this procedure were obtained from the Water Resources Information System (WRIS) supported online by Kentucky Infrastructure Authority (KIA). The final database of systems represents a revision and expansion of the database originally developed by Jolly et al. (2013). As before, systems were selected for the database based on spatial configuration and average daily demand. In order to protect the security of the utilities, all identifying information was removed such as names of pumps and tanks. Additionally, each model was given a generic name in the form KY #. Finally, as part of the revision, some of the systems were reclassified based on the more objective classification methodology proposed in this paper. In constructing each network model, a general schematic was created in ArcGIS for each of the systems using downloadable shapefiles from WRIS. The schematics were then imported to KYPIPE to create working hydraulic models. Elevation data was applied to each hydraulic model from downloadable digital elevation models, while maximum and minimum tank elevations were estimated based on bird s-eye view aerial imaging. Lastly, all systems were checked for connection errors using built-in software tools and by manual inspection. Figure 2.1. Model Creation Procedure.

3 Average daily demand data were taken from WRIS and distributed to individual nodes using a demand allocation tool which assigns fractions of the total demand to nodes based on diameters of the connecting pipes. This methodology for model development is illustrated using KYPIPE, but the general methodology will be applicable for most other commercially available software (e.g. Innovyze, WaterGEMS, EPANET, etc.). 3.0 WATER DISTRIBUTION SYSTEM CLASSIFICATION PROCESS Historically, water distribution systems have been classified as one of three basic configurations: branch, loop, or grid (see Figures ). Such classification is normally based on an subjective analysis of the configuration of the system as viewed from a graphical representation of the network (von Huben, 2005). While such a method can usually be successful in identifying branch systems, grid and loop system schematics often look very similar and are often undistinguishable without additional data (i.e. pipe diameters). Further, most systems tend to evolve over time and thus may have combinations of different types. Thus, an objective way to classify systems would be highly beneficial. Figure 3.1. Example of a Branch Network

4 Figure 3.2 Example of Loop Network Figure 3.3. Example of a Grid Network

5 In order to remove subjectivity from the procedure, the following methodology was used to classify each system in the database as one of the three basic configurations mentioned previously. All systems were initially assigned a configuration based on network schematics. Quantitative parameters (e.g. the ratio of branch pipes to total pipes, etc.) were regressed from the dataset to be used as the dividing line between configurations. The term branch pipe refers to any pipe not included in a loop. Furthermore, a 3-pipe loop is simply a loop containing 3 pipes while a 4-pipe loop is a loop containing 4 pipes, and so on. Based on an analysis of 15 systems, the following general rules were developed. Figure 3.4. Classification Algorithm. To automate this classification process, a simple visual basic code was developed that is able to read in a KYPIPE or EPANET input data file and then generate a histogram of pipe loop densities which can then be used in applying the classification algorithm (see Figure 3.5). Figure 3.5. Pipe Loop Density Histogram (for a grid system).

6 Once developed, the Pipe Loop Density Tool was applied to the 15 systems that were assembled as part of the network database. This resulted in a database consisting of 5 branch systems, 5 loop systems, and 5 grid systems. All statistical results presented hereafter pertain to the classification method used in this study. 4.0 STASTICAL ANALYSIS OF DATABASE SYSTEMS Once the database was finalized, differences in basic system characteristics such as number of tanks or average pipe diameter were investigated to determine if these characteristics vary systematically by configuration. Systematic variations found between configurations may be useful for several applications, the most basic being a simple classification tool which does not need a working hydraulic model. The following characteristics were investigated for this study: number of tanks, number of pumps, pipes to nodes ratio, total length of pipes, average pipe length, average pipe diameter, and the average number of pipes into or out of a junction. The statistics for each system were averaged with other systems of the same configuration. Hypothesis testing was performed on these averages to determine whether or not the configurations are statistically different; the null hypothesis being that the two values are statistically equivalent. A one-tailed two-sample t-test was performed for each statistic using a significance level of α = 0.05, implying 95% confidence in the decision. Results of the hypothesis tests are shown in Table 4.1. Significant differences were found in all parameters for branch to grid comparisons and branch to loop comparisons, and 3 parameters were found to be significantly different when comparing grid to loop systems. Table 4.1. Results of Hypothesis Testing. Configuration # of Tanks # of Pumps #Pipes/ #Nodes Total Length of Pipes (mi) Avg. Pipe Length (ft) Avg. Pipe Diameter (in) Avg. Pipes per Junction Branch & Loop B>L B>L B<L B>L B>L B<L B<L Branch & Grid B>G B>G B<G B>G B>G B<G B<G Loop & Grid L>G L=G L<G L=G L=G L=G L<G Confidence intervals were constructed for each of the system parameters by configuration and are shown in Table 4.2. A significance level of α = 0.05 was used, implying 95% confidence the statistic lies within the interval. Note that some confidence intervals may contain large overlaps for comparisons which did not reject the null hypothesis (e.g. the loop and grid comparison for number of pumps).

7 Table 4.2. Confidence Intervals for System Parameters by Configuration. System Parameter Lower Bound Branch Loop Grid Upper Bound Lower Bound Upper Bound Lower Bound Upper Bound # of Tanks # of Pumps #Pipes/ #Nodes Total Length of Pipes (mi) Avg. Pipe Length (ft) Avg. Pipe Diameter (in) Avg. Pipes per Junction SUMMARY AND CONCLUSIONS This paper has outlined a database of 15 water distribution systems models that reflect actual distribution systems in Kentucky. These actual networks provide researchers a robust database for testing new algorithms. The database is being provided free to the research community (via dropbox) with models available in both KYPIPE and EPANET formats. Access to the dropbox may be granted by ing the authors at wds@engr.uky.edu. This study also introduces a methodology for system classification not previously used. Systems in the database were further investigated to determine if certain characteristics vary systematically by configuration. Statistically significant differences were found when comparing characteristics for dissimilar configurations. These differences may be useful in helping to forecast system expansion needs (pipe, tanks, etc.), and security needs (i.e. number of water quality sensors, etc.) as the system continues to grow and expand. 6.0 ACKNOWLEDGEMENTS Funding for this research was provided by the U.S. Department of Homeland Security, Science and Technology Directorate, through a technology development and deployment program managed by The National Institute for Hometown Security, under an Other Transactions Agreement, OTA #HSHQDC , Subcontract #02-10-UK.This support was greatly appreciated. 7.0 REFERENCES AWWA. (2012). Computer Modeling of Water Distribution Systems. Manual of Water Supply Practices- M32. Denver: American Water Works Association.

8 Dielman, Terry E. Applied Regression Analysis. Mason, OH: South-Western Cengage Learning, Jolly, M. D., Lothes, A. D., Bryson, L. S., & Ormsbee, L. (2014). Research Database of Water Distribution System Models. Journal of Water Resources Planning and Management, Kentucky Infrastructure Authority. (2010, February 25). Water Resources Information System. Retrieved 2012, from Mays, L. W. (2000). Water Distribution Systems Handbook. New York, NY: McGraw Hill Mays, L. W. (2005). Water Resources Engineering. John Wiley & Sons, Inc. Murray, R., Janke, R., Hart, W. E., Berry, J. W., Taxon, T., & Uber, J. (2008). Sensor network design of contamination warning systems: A decision framework. Journal American Water Resources Association, McGhee, T. J. (1991). Water Supply and Sewage. Hightstown, NJ: McGraw-Hill, Inc. National Research Council. (2006). Drinking Water Distribution Systems: Assessing and Reducing Risks. Washington, DC: The National Academic Press. U.S. Environmental Protection Agency. (2008). Water Quality in Small Community Distribution Systems - A Reference Guide for Operators. Cincinnati, OH: Office of Research and Development. Von Huben, H. (2005). Water Distribution Operator Training Handbook. American Water Works Association. Denver, CO. Walski, T. M., Chase, D. V., Savic, D. A., Grayman, W., Beckwith, S., & Koelle, E. (2003). Advanced Water Distribution Modeling and Management. Bentley Institute Press.