Integration of large-scale wind power creates challenges for power system operations. One of the effective ways of dealing with these challenges is to build thermal power plants to form bundled wind-thermal generation system (BWTGS), i.e., using thermal power to alleviate the fluctuation of wind power. This paper presents a method for optimal capacity and type planning of BWTGS with the given wind farms. Branch-descending technique (BDT) is used to generate candidate schemes of thermal generating units by analyzing the rules of total cost changing with the reduction of the number of thermal generating units. The optimal scheme of BWTGS can be obtained by simulating a long-term operation process of BWTGS and comparing the total costs of all schemes. Techniques to accelerate computation, such as combining redundant states in dynamic programming (DP) algorithm and the saving-branch-cost technique in BDT, are developed to reduce the computational complexity. The major advantage of the proposed method is that it can be used to obtain not only the optimal capacity of thermal generating units, but also the optimal type and number of thermal generating units. Case studies are conducted to demonstrate the effectiveness of this proposed method. (C) 2015 Elsevier Ltd. All rights reserved.