The method to locate distribution networks fault is an optimizing problem with high-dimensional binary in essence based on information from sensor. In order to deal with possible that increasing of the number dimensions of the solution when distribution networks become large

and information loss such as sensor uploading and distortion of information inteference with algorithm optimizing causing failure of fault localization

present a new membrane system algorithm

which uses a nested structure that regards the bunch as a membrane structure to change number dimensions of the solution; besides

in order to balance contradiction between exploration and mining in optimization strategy

introduce chaotic strategy to improve fishing operators

so that chaotic fishing strategy can jump out the local optimum. Simulation results show that the proposed method can cary out singlepoint or multiple-point fault localization accurately

contrasting with genetic algorithm（ GA） and ELM algorithm

it has obvious advantages in convergence rate and accuracy

particularly suitable for large-scale distribution network

fault localization using real data for fault recurrence and the proposed algorithm for fault localization. Results show that the proposed algorithm can locate fault accurately

and carry out fault-tolerant processing on distortion data

which has practical meaning.