161 / 2022-07-05 23:42:35

Triplet Network for Topology Identification of Distribution Network

Triplet Network， PSO， GAF， Topology Identificatio

Power System and Automation > Smart grid and utility applications

The distribution network has the problems of inaccurate topology and incomplete measurement configuration. This paper proposes a  method of distribution  network  topology identification based on the triplet network. In order to improve the generalization ability of the model, the Latin Hypercube Sampling (LHS) method considering the source and load correlation  was  used  to  generate  PV  and  load  data.  A hybrid feature selection algorithm combining MLP and PSO is proposed  to  reduce  the  number  of  input  measurements.Sequence-to-image  conversion  using  Gramian Angular Field (GAF) is implemented to improve model training efficiency. We introduce a momentum encoder to select hard triplet samples, which  solves  the  problem  of  easy  gradient  dissipation  when triplet samples are selected randomly. The IEEE33 node system is used to verify the accuracy and superiority of the proposed algorithm, especially in small sample and weak loop network scenarios, the identification accuracy can reach 92% and 89%.