An improved planning model of tree-type irrigation pipe network is proposed with the purpose of minimizing the expenditure of construction. Pipeline connectivity is guaranteed by using the concept of high-level nodes. Meanwhile, an improved dual-frequency bat algorithm(DFBA) is proposed to balance the diversity and convergence of the proposed algorithm by dual-frequency strategy. The optimization 100 dimension results of 12 test functions indicate that the DFBA can enhance global search capability effectively in comparison to the genetic algorithm, the particle swarm optimization algorithm and the original bat algorithm. Finally, planning experiments are performed on 2 pipe network cases, which consist of 10 nodes and 34 nodes respectively. Compared to the original bat algorithm, the mean values of investment cost obtained using the DFBA can be reduced by 23.60% in the pipe network of 10 nodes, and 31.03% in the pipe network of 34 nodes, respectively.