Optimal Design of Sewer Networks using hybrid cellular automata and genetic algorithm

Optimal sewer design aims to minimize capital investment on infrastructure whilst ensuring a good system performance under specific design criteria. One of the state-of-the-art optimization techniques for this problem is the Genetic Algorithm (GA), which is commonly combined with a sewer hydraulic simulator during the optimization. However, this approach can be prohibitively time-consuming especially for designing large networks. Firstly, GAs normally take a large number of generations to achieve performance improvement. Secondly, many forms of GA rely on randomly generated initial populations which are often poor solutions. To overcome this intractable problem, this paper introduces a robust hybrid optimization method, named CA-GASiNO (Cellular Automata and Genetic Algorithm for Sewers in Network Optimization). It fulfils the design task at two stages. A local agent approach based on Cellular Automata (CA) principles is firstly applied to obtain a set of preliminary solutions, which are employed to seed a multi-objective Genetic Algorithm (MOGA) at the second stage for final polished designs. The CA based approach provides a good initial population at a remarkably small computational cost and hence saves computation for the following genetic algorithm runs. The GA targets the global optimal which is fundamentally troublesome to the localised CA approach. Two sewer networks, one small artificial network and one large real network, were used for case studies. All results indicate that the proposed method outperforms the standard multi-objective GA in terms of its optimization efficiency whilst achieving a better Pareto front.