Ultra high frequency (UHF) partial discharge (PD) detection is a common method for transformer oil-paper insulation defect location. However,the accuracy of PD positioning in practical application is easily affected by noise and sensor layout. In order to ensure the effectiveness of PD positioning detection of transformer oil-paper insulation defect,an UHF PD positioning detection platform for oil-paper insulation defects is established. Based on the conventional K-means method,a modified clustering demarcation PD anti-interference positioning method is proposed,which effectively reduces the positioning error. Aiming at the problem of sample clustering demarcation aliasing,when the optimal correction coefficient L is 1.1,the UHF PD positioning error can be reduced to less than 0.1 m,which verifies the effectiveness of the proposed method. Finally,the variation law of positioning error of different sensor layout is analyzed,and the optimal layout scheme of UHF sensor for transformer oil-paper insulation defect detection is proposed,which can provide reference for the layout and positioning of transformer on-line monitoring sensor.