In order to improve the substation station domain positioning accuracy of the ultra high frequency（UHF） method of partial discharge

we put forward a method of the correction of positioning error based on BP neural network

which can correct the measured values by constructing an error compensation surface in which r is from 0 to 60 meters and theta（θ） is from 0 to 360° degrees based on the training of error value of the finite point（radial distance r

azimuth angle θ）

and the K-Means machine learning algorithm is used to make the measured values debouncing. Moreover

taking the engineering application factors into account

we propose a method based on spline interpolation to obtain samples

which is feasible in project application. The experimental results show that the measurement error of radial distance rtends to decrease in the range of 0～40 m. The error of radial distance measurement increases within r>40 m

and the distribution of the direction angle measurement error is similar. After the error compensation

the error distribution tends to be stable and the error value is greatly reduced

which proves the feasibility and effectiveness of the method.