Abstract: Traditional partial discharge detection technology for power cable joints is mostly based on Beidou positioning and WT-SVM principle. However, in complex electromagnetic environments, these technologies have the risk of signal interference or misjudgment, resulting in significant deviations in detection results that cannot accurately reflect the actual situation of partial discharge. Therefore, we propose a research scheme for partial discharge detection technology of power cable joints based on improved wavelet packet. Firstly, using an improved wavelet packet decomposition method, the sub-band energy and energy entropy are solved as partial discharge features to reflect the energy concentration of partial discharge signals in different frequency ranges. Based on this, we use sensors for monitoring, collect data and process signals, and use cross-correlation algorithms to accurately locate discharge points, thereby achieving the detection target of partial discharge. The experimental results show that after the application of this technology, the strength of partial discharge signals detected by the cable at various time points is highly consistent with the actual values. When locating discharge signal points, this technology demonstrates extremely high accuracy with minimal deviation, almost completely matching the actual signal points. At the same time, it has the ability to accurately identify the discharge stage of power cable joints.