ABSTRACT
Models based directly on detailed hydraulic simulation have been used for scheduling (e.g. Powell et al., 1999), but they are inappropriate for linear programming and too slow for methods which require large numbers of schedule evaluations such as dynamic programming and metaheuristics. Macroscopic and linearised models based on regression have been used, for instance Lansey and Awumah (1994), but they are not “spatially responsive”, i.e. if there is a change in network connectivity or in the geographic distribution of demand they can be misleading (Ormsbee, 1991). A linearised model which can be built automatically and rapidly updated from a detailed hydraulic model should be both able to respond to frequent network changes and compatible with many different solution methods, including linear programming. This paper describes such a model, and explains how to build and update it.
