It is a long term goal for high voltage engineering field to acquire the discharge voltage of air gap accurately

and the research for discharge voltage prediction has important engineering significance. Based on data driven theory

the line-tower air gaps of 750 kV one tower double-circuit and 1 000 kV compact tower transmission lines were selected as training samples

and the electric field distribution of the gap was calculated. In order to characterize the structure of air gaps electric field features were extracted along the shortest path from the high voltage end fitting to the tower as input parameters. A discharge voltage prediction model of line-tower air gaps under standard switching impulse was built using support vector machine. Finally

the discharge voltages of UHV cup-tower outer phase air gaps were predicted. Compared with the experimental values in discharge characteristic test

the maximum error of predicted values with gap distances between 4 and 8 m is within 10%

while mean absolute error is 3.66%. The results verify the effectiveness of the prediction method. Based on discharge test data of various tower structures

the discharge voltage of other line-tower air gaps can be predicted by machine learning algorithm

which can provide a new way for the acquisition of discharge voltages in engineering gap.