Artificial Intelligence and Smart Grids: the Cable Joint Test Case

Abstract

The technological advance of the XXI century provides several benefits to the electrical grid. Such benefits are not limited to new electrical assets, also technologies developed for other fields may become key tools. A clear example is artificial intelligence (AI), which is fundamental to the processing of the data being generated nowadays. Therefore, a potential application of AI in the electrical grid is the predictive maintenance of medium voltage cable joints. These accessories are one of the main causes of fault in the distribution network, resulting in significant economic losses and energy not supplied to the customers. In this paper, a realistic scenario is designed to produce data for a typical machine learning (ML) algorithm. In detail, the main fault modes of cable joints and the associated parameters are defined. Afterwards, the ML algorithm is tested and validated considering its realistic implementation by a distribution system operator. From the results, it is possible to appreciate (i) the applicability and the effectiveness of the algorithm for the predictive maintenance of cable joints; (ii) the discussion on the pros and cons of the use of ML algorithms; (iii) some hints to better exploit the algorithm in practical applications.