Probabilistic weather forecasting for dynamic line rating studies

This paper aims to describe methods to determine short term probabilistic forecasts of weather conditions experienced at overhead lines (OHLs) in order to predict percentiles of dynamic line ratings of OHLs which can be used by a system operator within a chosen risk policy with respect to probability of a rating being exceeded. Predictive probability distributions of air temperature, wind speed and direction are assumed to be normal, truncated normal and von Mises respectively. Predictive centres are estimated as a sum of residuals predicted by a univariate auto-regressive model or a vector auto-regressive model and temporal trends fitted by a Fourier series. Conditional heteroscedasticity of the predictive distribution is modelled as a linear function of recent changes in residuals within one hour for air temperature and wind speed or concentration of recent wind direction observations within two hours. Parameters of the probabilistic models are determined to minimize the average value of continuous ranked probability score which is a summary indicator to assess performance of probabilistic models. The conditionally heteroscedastic models are shown to have appropriate sharpness and better calibration than the respective homoscedastic models.