Document Type


Publication Date

January 1985


Any water supply system can be conceptualized as consisting of three componenets--source development facilities (including treatment), transmission facilities, and a distribution network. The scope of this report is limited to the first two--the source related facilities upstream from the distribution network. In the mathematical modeling of regional rural systems, the number of variables, and hence the size of the model, increases rapidly as the number of system componenets and their alternative designs increase. Regardless of the method of solution, manual preparation of large models is cumbersome and is vulnerable to human error both in the computations of the matrix coefficients as well as in the format requirements. This research is aimed at developing a flexible matrix generator for the general rural water supply problem and alternative solution methods that can be used for especially large problems. The Mixed Integer Programming (MIP) method is particularly expensive to use for large problems and is not practical if the number of variables approaches a hundred or so, while the laternative solution methods can handle hundreds of variables. A real-world application problem is solved using the MIP and the three alternative methods developed during this research. These include the continuous, the nonlinear discrete, and the objective bounding methods. The solutions are compared and conclusions drawn as to the conditions under which the different methods are recommended.