In view of the current situation that the distribution network in China must be reconstructed from many perspectives in a limited number of investments， an adaptive genetic algorithm based optimization method is proposed to identity the reconstruction scheme in distribution network. In accordance with the vulnerable spots of the present network， several reconstruction projects and their relations are put forward while suitable encoding scheme and special genetic operators such as crossover and mutation are designed. Adaptive genetic algorithm is used to permute all the projects under the restriction of maximum investment and the combination with best comprehensive evaluation index is selected as the optimization result. The final example shows that the approach proposed can obtain satisfied result with various defects in and limited investment for the current distribution network.