Formal optimization analysis can help a designer find the best solution to complex distribution system planning, design and operations problems. Genetic Algorithm (GA) optimization recently was put to the test in the planning of a 12,000-home residential development in Arizona.
The conversion of vast areas of farmland to residential communities, industrial parks and shopping malls is a phenomenon occurring in many parts of the country. In most cases, the developments are relatively small and the developer submits a fairly "cookbook" water distribution plan to the local water utility, receives approval and begins construction. The water utility may or may not review the plan using a hydraulic simulation model.
In other cases, such as in the Arizona development, the developer needed to do some serious distribution system planning to come up with an efficient and cost-effective phased implementation plan. That meant hiring an experienced consultant to prepare a water system master plan and design the backbone transmission and distribution system.
The system must provide adequate high-quality drinking water and sufficient landscape median and golf course irrigation water as the developer constructs a series of subareas to reach buildout in six to 10 years.
The system includes a water treatment plant with associated wet well and main pump station, two booster pump stations, and a network of pipes to ultimately serve 12,000 homes in 13 subareas. As new subareas are constructed, storage facilities will be added to the system. The entire development will cover some 4,500 acres.
Planning for Buildout
Planning the water distribution system began with siting the water treatment plant and developing a trial layout of pipes and supply nodes for the buildout condition. Projected demands were prepared based on definite subarea plans showing housing units, commercial and industrial zones, and parks, golf courses and open space.
Planning for Buildout
The computed water demands were 9.36, 22.88 and 33.65 mgd for the peak hour, average day and maximum day. For the maximum day plus fire flow condition, an added 3,000 gpm was assigned to each node individually.
Planning for Buildout
The basic demand cases were analyzed using the following design criteria:
Planning for Buildout
- maximum pressure of 100 psi
Planning for Buildout
- max day minimum pressure of 40 psi
Planning for Buildout
- peak hour minimum pressure of 30 psi
Planning for Buildout
- max day plus fire minimum pressure of 20 psi
System Redundancy
One interesting feature of the backbone plan is that the consultant defined a system redundancy or reliability condition for two pipes crossing under the railroad tracks. If either of the pipes goes out of service, the other pipe alone must have enough capacity to supply the area south of the tracks with the average day demand plus fire flow.
System Redundancy
Using a simulation model, the consultant created and evaluated a series of trial solutions to find a feasible plan. Alternative pipe routes along the planned network of main roads were delineated. Pipe layouts were prepared with assumed pipe sizes and the trial solutions checked for the basic demand conditions and the system redundancy cases.
System Redundancy
Over many trials, the backbone plan was developed. The plan includes 34 supply nodes and 43 pipes ranging in size from 12 to 36 inches in diameter. The total length of new pipe is 97,856 ft. at an estimated installed cost of $7,129,000.
Optimizing the System
The consultant recognized that even for this small pipe network problem, the challenge of finding an efficient and cost-effective buildout plan was enormous. In fact, the total number of possible combinations assuming 11 pipe size choices (10-42 in. plus a zero size) and 51 pipe locations is 11 to the 51st power, which is an astronomical number.
Optimizing the System
The consultant requested a review of his proposed solution using OGA (Optimatics Genetic Algorithm) analysis. The OGA read in the simulation model data, the 51 allowable pipe locations, the 11 allowable pipe size choices, and the estimated installed pipe costs. A proper solution string format was developed and initial values of the genetic search parameters prepared.
Optimizing the System
Next, the critical demand conditions used by the consultant were input into the OGA model as constraints to be met in each OGA run. Maximum day plus irrigation demand and 19 critical fire flow cases were specified. The OGA then generated and evaluated several hundred thousand individual trial solutions as it searched for the lowest-cost combination of pipes that met each demand condition and all specified design criteria.
Optimizing the System
The OGA analysis quickly identified a handful of promising lower-cost alternatives. Three of the early optimized solutions came up with pipe costs ranging from $6,497,000 to $5,987,000 (9-16 percent less than the first design cost).
Optimizing the System
About the Author: Jeffery Frey, P.E., is president and founder of Frey Water Engineering, Inc., in Arlington Heights, Il. He oversees all GA optimization studies for the North American area.
Simulation Modeling Challenge
Frey Water Engineering is sponsoring a contest that pits simulation modeling against GA optimization. Readers who have access to simulation modeling software and can come up with the best solution for the below problem could win $500.
Simulation Modeling Challenge
The lowest cost solution of all feasible entries received by July 4, 1999, will win the prize. If there is a tie for first place, the earliest solution wins the whole prize.
Simulation Modeling Challenge
The challenge involves a simple system expansion problem aimed at demonstrating that simulation analysis is not really as easy as it might appear.
Simulation Modeling Challenge
Participants are given an existing network of 21 pipes supplied from one reservoir located at the north end of the system as shown. Working from a set of future demands that the existing system is incapable of supplying, contestants must locate and size new pipes to be laid parallel to one or more of the existing 21 pipes such that the upgraded network supplies the specified demands while meeting a minimum pressure of 40.0 psi at Nodes 2-19.
Simulation Modeling Challenge
The specifications are in the tables on page 32. For more information on the FWE/Optimatics simulation modeling challenge, visit the Internet site at: http://www.frey-water.com/challenge.htm. The page provides further details on the contest and allows participants to download the .INP and .MAP model files in EPANET format.
Simulation Modeling Challenge
In July, the company will report the winning simulation design and compare it to one or more GA-optimized solutions.
Simulation Modeling Challenge
NOTE: This challenge is only for simulation modelers- no optimization solutions, please. See www.frey-water.com for more details and to download a copy of the EPANET file.
