This article describes the partial discharges (PD) detection and localization system. Localization can be performed on the basis of measured UHF waveforms analysis during activity of the partial discharges. The time-shifts of the waveforms related to transient process occurrence in the signals are the main input parameters for localization methods. In order to estimate the position of the signal source in the 3D space a minimum of four antennas has to be used, since the time of the PD is unknown. Analysis of the PD starts with visual investigation of the signals present inside of the power oil transformer. PD signal level is in comparison to other signals inside the transformer very low. Therefore, signal line contains, except amplifiers, also controlled attenuators to reach required trigger conditions. Detailed investigation of all signals inside transformer is provided in memory mode in which system store defined number of signals for offline diagnosis a localization. Continual and memory mode of the system also contain discrimination method for non PD signals removing. Discrimination method is based on simultaneous detection of the signal inside and outside of the transformer with aim to acquire only signal which comes from inside of the transformer. Localization of the PD origin is based on TDOA and matrix algorithm. Resulting position are visualized as point in transformer 2D and 3D scheme.