The main goal of this thesis was to determine and characterize the MV underground cable behaviour of the Portuguese distribution network. This study is innovative in the area as there was no previous in-depth study on the matter. The objectives included the characterization of the network and determining the networks main traits regarding cable failures, which included accessing the matter of cable seasonality.  The outcome of the analysis carried out in this thesis may be used for distribution operator asset management decisions and also improvements for the data acquisition process and the communication between databases.

 

    This thesis provided the following contributions to knowledge:

 

     1. The importance of proper data acquisition for making reliable analysis. As mentioned throughout this thesis, the quality of the existing data limited the analysis and conclusions, as the missing information is vital for a survival or reliability analysis.

 

     2. Detailed network characterization. Analysing how the network is organized demonstrated how underground cables are concentrated in highly populated and urban areas. Additionally, this analysis showed that oil filled cables have been, over the year, replaced with dry or extruded cables. The lengths and quantities from this analysis were used for the cable failure analysis.

 

     3. Determining cable behaviour patterns. It was important when analysing the failure rates to determine if they were really caused by a high number of cable incidents or if they were a result of a low cable concentration, which means that failure rates are inflated. The regions with the highest cable concentration coincided with the highest failure rates. When comparing cables types, the oil filled cables had a worse performance, which can be explained by the fact that oil cables were older. In addition, analysis found that the worst performance voltage level was 10kV.

 

     4. Underground cables have a major impact on SAIDI and SAIFI QoS indicators. There has been a general improvement in the MV underground network over the period analysed. Analysing cable’s impact on the QoS indicators determined that cables increase the quality of service indicators such as SAIDI and SAIFI. The MAIFI indicator is not strongly affected by cable failures as it considers SD incidents, but the majority of cable incidents are LD, with an average duration of 161 minutes.

 

     5. Evidence for seasonal cable failure behaviour. While grouping the information in months and comparing with the average temperature of the month, there seems to be a high correlation between cable failures and temperature. But when this information is analysed with daily data, the evidence is very low, despite not being totally rejected. Also in terms of dependence, the result of Hoeffding’s test determined that it is very low. The hottest periods in the year, from April to July, seem to have a higher correlation, when comparing to the overall year and colder months. Maybe with more detailed and accurate data, this correlation accepted or rejected with statistical significance.

 

     6. Cable daily failure fitted to Weibull probability distribution. Analysing the number of daily incidents, excluding null incident days, a distribution was fitted to the data according to the MLE method and the Kolmogorov-Smirnov goodness-of-fit test. The Weibull was determined as the best fit, despite the hypothesis for other distributions to fit the data could not be rejected. The fitted Weibull CDF can give a conditional probability of the number of cable failures for a certain day. So, knowing that there is an incident, it gives the probability that a certain number of failures will occur. This distribution can be updated every year with new cable failure information.