When the City of Grand Prairie, Texas, Utility Services Department issued its request for proposals for a Comprehensive Water Distribution Master Plan, the RFP included the following language calling for the use of genetic algorithm analysis.
The year was 1997 and the particular RFP language specifying that genetic algorithm (GA) optimization be used caught a lot of people by surprise. The Utility Services Department received numerous phone calls from consultants asking: What is genetic algorithm analysis?
Director of Utility Services Ron McCuller was the person responsible for including the language. McCuller had learned about GA optimization in April when he stopped by the Frey Water Engineering booth at the Texas AWWA Conference in Arlington.
As McCuller later explained, he wanted a method of providing confidence that not only were the proposed improvements necessary to provide for future growth and improve current operational problems (low pressure and water quality complaints), but that they be the most cost-effective solutions.
What McCuller found in researching GA was that this formal optimization technique had the potential to develop cost-effective distribution system solutions. The team of Optimatics/Frey Water Engineering had completed a Water Master Plan review for Fort Collins-Loveland Water District in Colorado in 1995. FCLWD's recommended plan to meet year 2015 demands called for 46 new pipes totaling 29.4 miles.
The Optimatics GA (OGA) review evaluated alternative connections to existing and proposed supply sources and different mixes of pipe diameters. The OGA-optimized solution had just 22 new pipes that totaled 18.8 miles, while meeting the same demands and design criteria. The estimated capital cost of the system expansion dropped from $5.85 to $2.97 million. FCLWD has now adopted the optimized plan as the basis for the District's new capital improvement program.
The aim of Grand Prairie's Compre-hensive Water Distribution Master Plan was to develop a new hydraulic simulation model, then use the model as the basis for preparing a system Master Plan, including phased capital improvement plans for future years. The alternatives evaluation phase of the study would utilize GA optimization to ensure that superior, low cost solutions were identified. The GA would optimize not only capital improvements, but also operational decisions.
The city decided to insist on optimization for a number of reasons. First, the Grand Prairie service area has a long, narrow configuration (26 miles by 4 miles).
A large lake divides the service area, nearly spanning the city's full width. These features present some difficult (and expensive) utility service problems, especially in extending service to the largely undeveloped southern areas where explosive growth is forecast in coming years.
Finally, the city had a number of operational problems and it was expected the optimization could identify cost-effective solutions. The GA would investigate options to reconfigure the city's four pressure planes by opening existing closed pipes and by adding new inter-connections between adjacent high flow mains. The GA would also consider options for abandoning an existing pumping station and storage tank to simplify operations and improve system performance.
Following a competitive selection process, the team of CH2M Hill and Optimatics/Frey was awarded the Grand Prairie Water Master Plan. CH2M Hill would first develop a system-wide hydraulic simulation model by extracting data from the city's water system design files.
CH2M Hill also prepared growth and demand projections to year 2025. System performance criteria were established, potential future water sources were determined, and possible capital improvement elements and operating strategies were identified.
Once the hydraulic model took shape, Optimatics/Frey began to formulate the OGA model and initiate the alternatives evaluation to identify several Master Plan solutions. A projected year 2015 demand level was adopted for the analysis. A series of preliminary, interim and final OGA runs were conducted to develop a range of feasible low cost solutions for consideration by the city. After a preferred solution was selected, CH2M Hill proceeded to finalize the Master Plan and detail the phased capital improvement plans.
Overriding GA Study Decisions
Decisions regarding supply rates from different sources and siting of new storage were critical to Grand Prairie. Because its average day and maximum day demands were projected to rise from 15 mgd and 30 mgd today to about 51 mgd and 92 mgd in year 2015, the city needed to carefully plan for its future sources of water.
The OGA optimized supply rates from three or four sources: via the existing Fort Worth, North Dallas and South Dallas connections, and the option of adding a fourth connection to bring in Midlothian water from the south. Related to this choice, the city also needed to locate 47 million gallons of new ground and elevated storage in the system.
A Challenging GA Problem
The Grand Prairie Master Plan optimization turned out to be quite a challenge. The system hydraulic model had 2,100 pipes, 1,830 nodes, 4 source connections and 11 wells, 19 pumps at 8 pump stations, up to 26 storage sites, and 10 flow control and pressure reducing valves. Each OGA run evaluated a maximum day extended period simulation (EPS) while optimizing more than 300 decision variables. The OGA had as its objective function to minimize project life cycle costs, namely:
- the cost of new transmission mains
- the cost of new & parallel distribution pipes
- the cost of new ground & elevated storage
- the present value of water purchase costs from 3-4 sources having different unit rates
- the present value of pumping energy costs
A series of final OGA runs optimized five distinct scenarios based on different assumptions for water supply sources and new storage locations (as suggested by the city). The resulting five optimized solutions gave the city decision-makers a real choice. Capital costs for the year 2015 solutions ranged from $66-$93 million, while lifetime operating costs ranged from $295-363 million. In general, the lower the capital cost, the higher the annual operating cost.
