Improvement of transmission line ampacity utilization via machine learning-based dynamic line rating prediction

Abstract

Transmission system operators operate overhead lines in power transmission networks by using thermal ratings calculated under static conditions. These static assumptions sometimes lead a network to work outside the range of safe conditions, and sometimes to work underutilized. For this reason, the use of dynamic ratings, which depend on the meteorological conditions of the region under study and thus are more adaptable and better able to ensure optimal operation, has become common. The main drawbacks of these dynamic rating calculations are that to perform day-ahead network scheduling, the ampacity must be known in advance, and unlike static ratings, dynamic ratings are complex to predict due to their great variability. This work defines a methodology based on machine learning techniques that enables the prediction of the ampacity of overhead transmission lines to facilitate the adjustment and optimization of the amount of energy that can be safely transmitted through a network. The results have been validated with real data gathered by sensors from an overhead line. In conclusion, the safety and working conditions of power lines can be improved by applying the selected models, since the number of periods working out of safe conditions can be reduced approximately from 18 % to 5 %.