Abstract: The cost of offshore wind farm's collector system constitutes a significant proportion of the wind farm's total investment, it is thus very significant to make a topology optimization of the collector system for reducing the fixed investment. The topology optimal problem can be modeled as a minimum spanning tree problem with dynamic edge weight. Because the edge weight is coupled with topology optimization, it cannot be solved by traditional methods. Through improvement in initial population selection, use of linked list coding and elite genetic operators, the improved genetic algorithm is engaged in the paper, which can not only improve the algorithm's efficiency, but also better address the complicated restrains of uncrossed marine cables. The case study shows the improved GA has good capability in optimization searching and convergence.