Abstract: In view of the fault recovery of complicated power distribution networks，a load transfer method for power distribution networks is proposed based on the graph theory and the improved fuzzy genetic algorithm． First，the to-pological structure and component parameters of feeders are integrated together by means of the JGraphT － based da-ta modeling of power distribution networks，and by taking advantage of the topological structure of power distribution networks explicitly，the tedious node encoding rules and the storage modes in the form of miscellaneous adjacency lists or matrices are avoided． Next，with the help of the graph theory algorithms integrated in JGraphT，different types of fault sections are quickly distinguished and the topological constraint is rapidly determined． Then，a for-ward － backward sweep method based on the recursive graph theory is developed． Finally，an improved fuzzy genet-ic algorithm is proposed according to the characteristics of power distribution networks． In the proposed algorithm，the initial solutions and genetic operators of the genetic algorithm are dynamically adjusted so as to improve the opti-mizing performance of the proposed algorithm，and the fuzzy inputs and fuzzy rules are revised reasonably to im-prove the convergence rate of the genetic algorithm and avoid the premature convergence． Simulation results show that the proposed algorithm is superior to the other methods in terms of power flow calculation and optimization effi-ciency．

Key words: JGraphT, load transfer, forward-backward sweep method, improved fuzzy genetic algorithm