Winding distortion is a common faults inside the transformer. The traditional mature detecting method of winding deformation belongs to off-line detection

and can not judge the winding deformation mode. According to the above reasons

this paper proposes a detecting method of transformer winding deformation based on distributed optical fiber sensing. The built-in distributed optical fiber continuous winding model is used to simulate the winding deformation in practical operation. When the winding is partially deformed

the optical fiber strain will be measured by Brillouin optical time domain reflectometer（BOTDR）. At last

the extreme learning machine（ELM） will make mode recognization to the detection signal. According to experimental results

the distributed optical fiber has some prestress in the winding

and the variation of fiber strain curve corresponds to different winding deformation. The accuracy of ELM is more than 90%for the winding and different deformation forms. The distributed optical fiber sensing technology can effectively detect the transformer winding deformation

which-provides a new idea for on-line monitoring of transformer winding deformation.