A multi-objective and multi-parameter optimization is implemented to design the optimal structure of bismuth-telluride-based TEG (thermoelectric generator) module. A multi-physics TEG model combining the SCG (simplified conjugate-gradient) algorithm is used as the optimization tool. The semiconductor pair number, leg length, and base area ratio of semiconductor columns to TEG module significantly affect the TEG performance, and hence are all incorporated into the present optimization study. A single-objective optimization is first implemented to provide input parameters for the multi-objective optimization. The results show that when taking the output power as the single-objective function, the output power can be elevated significantly by optimization of the three geometric parameters but which also accompanies the significant reduction in the conversion efficiency. The same result also occurs when taking the conversion efficiency as the single-objective function. By combining the output power and conversion efficiency with a weight factor as the multi-objective function, the optimization is again implemented. The optimal design obtained by multi-objective optimization makes a proper balance between the output power and conversion efficiency, so that the both are improved simultaneously. (C) 2014 Elsevier Ltd. All rights reserved.