Abstract: In this paper,a new kind of model for optimization of pumping tree pipe network and its solving method by Multi-Group genetic algorithms were proposed, in which the annual cost was used as an object function, the length and diameter and pump head were used as optimum variables. In order to keep the astringency and the variety, the fitness function is the sum of the object function and simulating annealing penalty function, and the cross generational selection technique is adopted in this algorithms.The simulation results indicate that the performance of the model and algorithms on pumping tree pipe network is excellent.