Abstract Partial Discharge (PD) is the most important phenomenon in deterioration of XLPE cable insulation. Then, the PD detection and especially online PD monitoring is an important idea for MV or HV cable networks protection, where are developed recently. Of course, the proper PD detection needs suitable noise reduction procedure and an appropriate algorithm for diagnosis of multi and simultaneous PDs that are generated by not only different defects along the protected cable, but also adjacent equipments. Moreover, the PD signals should be obtained without distortion due to intermediate equipments. High frequency inductive and capacitive electrical sensors which are used to current and voltage PD pulses, respectively, have related transfer functions; therefore, the output of any sensor is frequency dependent that applies a distortion on the primary signals. In this thesis: 1. First, a comprehensive lumped model for PD is proposed which is able to show and justify the PD treatment due to the changing of not only the voltage amplitude, the power frequency and the temperature, but also the shape, size and the location of the cavity inside the cable insulation. This model can also consider the electrical field distortion in adjacent the cavity. Based on this model, a novel differential model is proposed to show the transient behavior of non-homogenous and distorted electrical field due to multi cavities in the cable insulation. 2. Second, the frequency dependent and distributed model of the XLPE cables is analyzed. The attenuation and distortion of PD signals, when travel along the cable, are studied; and a novel algorithm is proposed to reconstruct PD current signals in the related defect point. The proposed algorithm is based on travelling wave theory and is applicable to simultaneous defects. 3. Third, the sensors distortion is analyzed and a new method is introduced to compensate the related distortion in online conditions. Compensation is done using the inverse of sensor transfer function, and therefore, the PD signals can be obtained as the similar to the original waveform. The performance of the proposed method is approved, experimentally. 4. A novel filtering procedure for extraction of the PD signals through the ambient noises, as the biggest problem in online PD detection, is proposed as the fourth step. A new wavelet transform based filter is constructed to decompose any primary non-stationary signal in any arbitrary frequency spectrum, without need to use previous approximate components with very low calculation burden (less than 50%). Also, an online adaptive noise reduction procedure is developed that is able to follow the ambient noise levels and adjust the threshold levels in any frequency spectrum, separately. Moreover, a mixing procedure to reconstruct broadband filtered components is proposed. 5. As the final and the main step, an algorithm for online location of simultaneous PDs is proposed, based on travelling wave theory. This algorithm uses the pulses train (generated by defects) along the protected cable. Both elimination of the fundamental frequency component and discrimination of PD pulses with reverse polarities are done using Mathematical Morphology idea. The simulation results show the high accuracy (more than 99%) for multi PDs locations even in noisy conditions. | 
