Optimal Operation of Raw Water System Using Accelerating Genetic Algorithm

Abstract

Security of urban water supply is threatened for lacking of enough water resources of coastal cities in China. In order to guarantee the safety of supplying raw water many reservoirs and pipelines are constructed. So complex raw water systems are formed, which include many reservoirs, constant-rate pumps and variable speed pumps. But the cost of transporting raw water is often higher because the systems are operated by expert’s experiences mostly. Genetic algorithm (GA) is used for solving combinatorial optimization problems widely, such as the optimal operation model of raw water system. But GAs have been criticized mainly for their premature convergence, enormous calculation cost and shorting of criteria of determining parameters. As an improved GA, accelerating genetic algorithm (AGA) could alleviates these difficulties. Firstly, parent population is generated randomly in given ranges. After taking selection, mutation and variation, top individuals will be found. Then the ranges of the variables in the individuals can be recorded and set as the initial ranges of the variables in the child population. Secondly, child population is generated randomly in the ranges. The population will be taken the same operation, so the ranges of top members will be adjusted and contracted gradually in evolution. The search space will be reducing gradually so that AGA could find an optimal solution easier and more quickly. At last the probability to acquire optimal solution will be increased. Convergence speed and the efficiency of global optimization of GAs are improved after applying these measures. This proposed method has been tested in the raw water system in ZhuHai. These results indicate that AGA outperforms the traditional GA and its empirical method in finding optimal solution is quicker and more efficient.