Improved Genetic Algorithms For The Optimization Of Urban Water Networks

The accomplishment of this paper is concerned with the practical water networks upgrading and expansion engineering. Urban water network is an important part of the water-supply systems. On one hand, urban water network is a bridge linking water-works with consumers. On the other hand, water network accounts for 50～70 percent of total water-supply systems budget, and takes about 30～50 percent of energy-consumption of total water-supply systems. Therefore, it is essential to optimize the water network for reducing the paving and operation cost and for improving the competence of water utilities. The optimization of water networks must meet the hydraulic constrains. So it is necessary to combine the optimization of water networks with the hydraulic analysis, and the performance of hydraulic analysis affects the optimization of water networks directly. Thereby analysis of the optimal model and hydraulic conditions of water network is important. On the basis of the existing achievements, this paper presents an algorithm by combining the square-root algorithm with the node-analysis algorithm. Application of the algorithm illustrates that it is effective and the performance is acceptable.Through the analysis and comparison of the most-used optimal algorithms, this study selects the genetic algorithms to optimize water networks. At the same time, the author describes an improved genetic algorithm to resolve the discovered problems of the simple genetic algorithm. Applications of the algorithms at the practice show the improved genetic algorithm is far better than the simple genetic algorithm no mater at the results or at the runtime. Comparison of the optimal solution and the normal solution of water network upgrading and expansion engineering in some city of Si Chuan Province illustrates the optimal plan which applies genetic algorithm saves the cost of water networks paving compared to the convention plan.Besides, the author analyses the capability of genetic algorithm and brings forward the recommended range of parameters. At the last part of this paper, the author also recognizes the future work that for us to do.