This project, carried out in collaboration with a complementary research team at Glasgow Caledonian University, developed our understanding of electrical discharges in the high voltage (HV) cables that are used in electricity distribution networks. Partial discharges (PD) within the insulation (usually at joints and terminations) erode material and are one of the more common causes of cable failure, which can lead to loss of supply and considerable inconvenience during repair. The research carried out involved measurement of the pulses of electrical current emanating from such defects in the early stages of their development. Since they can only be detected by sensors situated at the ends of cables (often hundreds of metres away), understanding how the pulses propagate and become distorted due to the cable material properties is important for implementing effective diagnostic techniques. Advanced high-frequency electromagnetic modelling software was applied to simulate propagation through the cable once the cable materials had been properly characterised. This model implemented the local excitation of a propagating pulse in the cable, and was combined with the propagation equations for long distances (based on measurements of cable samples) and the measured response of sensors (current transformers). By this means, the process has been fully modelled in a way that allows for better interpretation of the remotely measured signals to relate them to the severity of defects in the cable. Sections of defective 11 kV cable removed from service by an industrial partner were subjected to extensive forensic evaluation, and used to validate several forms of defect identification. Time-of-flight and thermal imaging techniques were used to locate the defects, after which computer aided X-ray tomography and scanning electron microscopy were applied to characterise the defect and identify manufacturing errors in the insulation system. Artificial defects created in a section of EPR cable allowed PD signals to be measured, characterised and added to a library for analysis using the newly developed classification algorithm. Two algorithms for knowledge discovery (Rough Set and Kohonen mapping) were evaluated and the Rough Set method was found to be superior in terms of its ability to interrogate and learn from existing databases, analyse PD activity from on-line monitoring systems, with the ability to evolve continuously and produce knowledge rules on the level of degradation. A novel (K-means) method was applied for automating recognition of PD patterns recorded under conditions in which a phase-reference voltage waveform from the HV conductors is not available. The method has proven particularly useful as it is capable of recognising patterns of PD activity in on-line monitoring applications for both single- and three-phase cables and is also effective for rejecting PD-like interference signals. A new, portable cable PD monitoring system was developed and has been applied to in-service cable PD tests at 4 large power stations. An insulation defect localisation algorithm, based on pulse rise time degradation with distance, was also developed from experiments and the on-site measurement campaign. This allows the system developed to provide an indication of the source and origin of PD.