Overhead Line Ampacity Forecasting With a Focus on Safety

Abstract

Predictions of the ampacity of overhead lines can be framed in a general context that aims to make electric grids highly efficient and reliable. In this paper, a methodology is presented that provides ampacity forecasts, which are valid for both the very short term, such as a few minutes or hours, and longer terms, up to 24 hours ahead. The former can be useful for grid operations, while the latter may be valuable in electricity markets. A time series methodology and mesoscale weather forecasts have been combined in machine learning algorithms for producing reliable ampacity forecasts for a span located in complex terrain. In a prior step, the developed algorithms made point forecasts, but finally, a computationally inexpensive algorithm produces probabilistic forecasts. These probabilistic forecasts are nonparametric, as they are not based on predefined probability distributions, and they demonstrate how a low risk in overhead lines is closely related to the reliability of ampacity forecasts.