Article “Probabilistic Forecasting of Dynamic Thermal Line Rating with Temporal Correlations” was accepted to the International Journal of Electrical Power and Energy Systems. Congratulations Tomas!
Abstract: Dynamic Thermal Line Rating is a technology that optimizes the utility of overhead power transmission lines by dynamically adjusting the rating according to current ambient conditions. It is often applied in discrete time intervals (i.e., the rating is updated hourly or daily) based on a rating forecast for the next period. This paper examines the effects of temporal discretization on rating prediction, discusses the importance of temporal correlations, and proposes an approach to include these correlations in a rating prediction system. It demonstrates that predictions may be biased towards extreme values if temporal correlation is not taken into account. The proposed solution to this problem is a sampling algorithm based on iterative sampling from a chain graphical model to emulate the correlations as observed in real data. This results in a set of time series samples with well-calibrated joint probability between neighbouring variables. Therefore, these samples are usable in a Monte Carlo optimization of an objective function and yield good calibration overall. The experimental evaluation shows that the dynamic rating calculated through this sampling method permits the usage of the result at its face value without the need for further calibration. It also increases the average daily rating compared to the basic quantile selection method.