부산대학교 도서관

In urban regions, where distribution electrical networks tend to rely highly on underground cables, operational reliability indices and maintenance costs strongly depend on the associated cable joint failure rates. As witnessed over the years, climate change has been increasingly contributing to such failures. Based on the field experience compiled systematically by Dubai Electricity & Water Authority distribution networks, this paper studies the seasonal patterns and the impact of various weather conditions on medium voltage underground cable joint failure rates. Among a list of commonly reported weather parameters, the most relevant of them are selected in terms of their impact on cable joint failures. Then, failure data are clustered according to the prominent features and the relative positions of the clusters representing the criticality of the corresponding weather conditions on the underground cable systems, which can be further quantified by predictions on failure rates. The predictive models use historical failure rates combined with the weather data to estimate future failure rates. Then, analyzing the performance of four different machine learning models and assessing the relationship between prediction errors and weather events, we extract useful insights for integrating weather and seasonal data in distribution system operation and planning procedures. One novel aspect of this study lies in geographical area (Middle East) and environmental conditions pertinent to the distribution network, as opposed to previous studies representing European countries and climate zones. Moreover, this work introduces a methodology to assess the association between rainfalls and cable joint failure rates.