A methodology for improved operational optimization of water distribution systems (1998-2001)

Funding body: Commonwealth Commission

Operational energy costs make up a substantial proportion of the annual expenses of water supply utilities. It is thus important that the operational control of water distribution systems is optimized to ensure that appropriate levels of service and reliability are met at minimum cost. The operational optimization problem is complicated by a number of factors: vast numbers of possible operational solutions; variations in demands and electricity tariffs through a typical operating cycle; minimum reservoir level requirements; and limitations on the number of pump switches. An additional complication is the non-linear hydraulic behaviour of water distribution systems, which makes computer modelling of these systems computationally expensive. One of the most effective ways of optimizing the operation of water distribution systems is through the application of genetic algorithms (GAs).

The GA methodology is based on the mechanics of natural selection, combining survival of the fittest with randomized information interchange between the members of a 'population' of possible solutions. A number of studies have shown that GAs can be successfully applied to the operational optimization of water distribution systems. One of the greatest drawbacks of GAs is that they require a large number of function evaluations to achieve convergence. Each function evaluation requires a computationally expensive dynamic simulation of the distribution system which, in turn, makes GA runs time-consuming. The objective of this project was to improve the efficiency of GA optimization. This was achieved by improvements in two areas. The first improvement was made in the dynamic modelling of water distribution systems. A new method, called the Explicit Integration method, was developed. In the Explicit Integration method, the system's hydraulic behaviour is linearized, and the reservoir demand described by a general polynomial function. This allows the reservoir dynamic equations to be solved explicitly. The linear hydraulic coefficients are updated by performing snapshot simulations at regular intervals during the run. The number of snapshot simulations is significantly less than those required by the conventional Euler numerical integration method. By reducing the number of snapshot simulations required for a dynamic simulation, the computational effort is reduced, and thus the simulation running time.

The second improvement to the efficiency of GA optimization was made by combining the GA with a local search method, the Hooke & Jeeves pattern search, in a hybrid method. GAs are good at finding the region of the optimal solution in a large solution space, but much less efficient in then finding the optimum point. Local search methods, on the other hand, are efficient in finding a local optimum, but are not able to escape the attraction basin of this point to search the wider solution space. By combining the GA and local search methods, the advantages of both methods are exploited to improve the efficiency of the optimization method. A number of example applications are used to illustrate the workings of the different methods developed in the study. In the final chapter, both the Explicit Integration and the hybrid methods are applied to a large and complex water distribution system in the UK. It was possible to reduce the time required for an operational optimization run substantially from a number of days to approximately one hour.

References

  • Van Zyl, J. (2001) A Methodology for Improved Operational Optimization of Water Distribution Systems, PhD thesis, University of Exeter.
  • Van Zyl, J., D.A. Savic, G.A. Walters (2002) Operational Optimization of Water Distribution Systems Using a Hybrid Genetic Algorithm Method ASCE Journal of Water Resources Planning and Management, (submitted for publication).
  • Van Zyl, J.E., D.A. Savic and G.A. Walters (2000), A Method for Improved Efficiency in the Dynamic Modelling of Hydraulic Networks, Second ISSMO/AIAA Internet Conference on Approximations Fast Reanalysis in Engineering Optimisation, May 25 ? June 2 (proceeding published on CD), p. 10.
  • Atkinson, R.M.A., van Zyl, J.E., G.A. Walters and D.A. Savic, (2000) Genetic Algorithm Optimisation of Level-Controlled Pumping Station Operation, Water Network Modelling for Optimal Design and Management, CWS 2000, Centre for Water Systems, Exeter, UK, pp. 79-90.
  • Van Zyl, J.E., D.A. Savic and G.A. Walters (2001), An Explicit Integration Technique for the Dynamic Modeling of Water Distribution Systems, World Water & Environmental Resources Congress, May 20-24, Orlando, Florida, edited by Phelps, D. and G. Sehlke (proceeding published on CD), p. 10.
  • Van Zyl, J.E., D.A. Savic and G.A. Walters (2001), "A Hybrid Method for Operational Optimization of Water Distribution Systems", in Water Software Systems: Theory and Applications, Vol. 2, Ulanicki, B., Coulbeck, B. and Rance, J.P. (eds.), Research Studies Press, Baldock, Hertfordshire, England, pp. 89-97.
  • Van Zyl, J.E., D.A. Savic and G.A. Walters (2002), Accuracy Issues in Extended Period Modelling of Water Distribution Systems, First Symposium on Environmental and Water Resources Systems Analysis, ASCE, May 19-22, Roanoke, Virginia, Kibler, D.F. (ed.), (proceeding published on CD), p. 10.

Back to Water distribution system management

Google+