In order to solve that Partial Discharge (PD) localization method based on time delay requires high level device hardware and is hard to promote, a PD localization method based on fingerprint spectrum of Ultra-High Frequency (UHF) signal feature parameter is proposed. This method includes off-line stage and on-line stage. In the stage of off-line, the test scheme aimed at test space is designed, the feature parameter fingerprint spectrum is established finally after data pre-processing and feature parameter extraction. In the stage of on-line, the genetic algorithm and least squares-support vector machine are combined for realizing the fast and accurate PD localization. The proposed localization method is tested through simulation test, the average localization error is 0.237 m, 92% localization errors are less than 1 m, and the variance of localization results for 20 times is 0.226. The test results show that the proposed localization method is of accuracy and stability. This work is supported by National Natural Science Foundation of China (No. 51877126) and Science and Technology Project of State Grid Shandong Electric Power Company (No. 2018A-058).