At present, it is difficult to obtain the global optimal solution fast in the large-scale transmission networks planning solving. For that, this paper proposes a model of transmission network planning which takes line investment costs, network loss cost, the over-load cost of normal operation, transmission corridor cost and N-1 overload penalty cost constraints as objectives. After the main disadvantages of being inclined to premature convergence, low accuracy and convergence rate in harmony algorithm are discussed, with the expansion of the transmission network and the increase in the dimension optimization problem, chaos adaptive grouping harmony search algorithm is proposed. Through group dimension, mutation and chaos disturbance to improve search ability of the algorithm, it can quickly solve large-scale transmission networks planning problems. Through the results of 18-node system and Southern Brasilian 46-bus system, it can not only prove the algorithm and model correctness and effectiveness in transmission network planning, but also can demonstrate that the algorithm has high computing speed and convergence, which provides the foundation for practical engineering applications.