The global demand for fresh water increases every year, while there is less and less of it available from natural sources. One way to produce more supply is with desalination of seawater. This can require pipelines to move water from the desal plant to the cities where it is needed. The complex operations of large water tank-and-pipeline networks present a good opportunity for numerical optimization. While during normal usage these networks can be expected to more or less follow a daily cycle, an optimizer is helpful when planning maintenance, when dealing with equipment failures or supply interruptions, and in planning extensions to the pipeline network.
The water optimization problem is somewhat easier than optimizing a completely general liquid pipeline: there are no batches; all tanks in the system are used for water; and thermal effects can be largely ignored. This makes it feasible to perform complete end-to-end optimization of the system, solving for tank levels, pump and regulator valve setpoints, pressures and flow rates to meet the demand schedules while obeying all operational constraints and minimizing the total power and maintenance costs. Operational constraints include pump and tank out-of-service for maintenance as well as supply interruptions.
This paper presents such an optimization approach. This optimizer uses multiple stages to solve different parts of the problem, applying a combination of nonlinear programming, dynamic programming, simulated annealing, and a transient hydraulic model. Our scheme takes advantage of the unique problem simplifications possible in a water network. We present the algorithm and some example scenarios on a pipeline network typical of the real-world networks where Atmos International has installed this system. We will also present some of the challenges that arose during the testing and commissioning process.
1.1 Water Pipelines
Long-distance water transmission pipelines are needed when the source of water for a city or region is distant from the city itself. The southern end of the California Aqueduct that provides water to Southern California from the snowpack in the Sierra Nevada Mountains is an example, with its southern half a pumped pipeline section. The largest such pipelines are in Saudi Arabia, pumping water from immense desalination plants to the cities. Libya’s Great Man-Made River is another example, irrigating all of Libya through a network of pumped pipelines sourced from the world’s largest fossil aquifer.
The largest of these pipelines have multiple pump stations and are comparable in complexity to a mid-sized liquid hydrocarbon pipeline. This suggests that their operations are good candidates for numerical optimization.