The tasks of high voltage and large capacity power transmission undertaken by high voltage cross-linked cable line are complex. Aiming at the problems that the detection accuracy of line oscillating wave partial discharge is poor and it is difficult to locate defects, a high voltage cross-linked cable line oscillating wave partial discharge detection method based on long short-term memory (LSTM) network algorithm is proposed. The oscillating wave voltage method is applied to establish the framework of oscillating wave partial discharge detection of high voltage cross-linked cable line. Based on the wavelet packet decomposition algorithm, the characteristics of typical partial discharge signals are extracted. The oscillating wave partial discharge signals are identified and detected by LSTM network algorithm to eliminate the noise in the partial discharge signals. According to the time difference between the original oscillating wave and the reflected oscillating wave reaching the test end, combined with the propagation speed of oscillating wave, the defect location of high voltage cross-linked cable line is determined, and the partial discharge detection of cable line oscillating wave is realized. The test results show that the partial discharge signal recognition accuracy obtained by the proposed method is higher, and the cable line defect location is more accurate, and the practical application performance is better.