An optimization model for gravity treetype pipe network is established, in whi ch the minimal investment is taken as the objective function, and the pressu re, flow rate and pipe diameter are taken as the constraint conditions. The impr oved genetic algorithm is used to solve the problem. Based on the optimal featur es of treetype pipe network, the dual coding combining binary coding with inte ger coding is adopted in the genetic algorithm to optimize the pipe layout and p ipe diameter simultaneously. Based on characteristics of tree in the graph theor y, the cycle method is adopted to improve cross method in the operations of init ial solution creating and mutation so that the number of infeasible solutions is reduced. Some operational processes of genetic algorithm are improved. The simu lated annealing algorithm is introduced in the model. The fitness function is ad justed, and the computing methods of crossover rate and mutation rate are improv ed. Example shows that the algorithm is efficient.