The reconfiguration of distribution network can reduce the loss of positive power in normal condition. There are many optimization algorithms for the reconfiguration of distribution network. As an emerging algorithm, Teaching-Learning-Based Optimization (TLBO) Algorithm has some advantages. For example, the algorithm has less algorithm-specific parameters, and is easy to understand and converge quickly. So TLBO is very suitable for the reconfiguration of distribution networks which is a multi-objective and multi-constraint problem. This paper chooses the minimization of the net loss and the switch operation times as the objectives, considers the costs of operations, and introduces TLBO into the research of distribution network reconfiguration. It elaborates on the coding strategy and the principle of modification of individual during iterations and self-adaptive modification of teaching factor in TLBO algorithm. The simulation results of PG&E 69 and a practical distribution network of a city validate the effectiveness of the improved TLBO proposed. This work is supported by Research Fund for the Doctoral Program of Higher Education of China (No. 20110141110032).