Dynamic Line Rating Forecasting Based on Integrated Factorized Ornstein–Uhlenbeck Processes

Abstract

Due to the intermittent nature of dynamic line rating (DLR) of overhead lines, DLR forecasting plays an important role in the scheduling of power networks. In the DLR forecasting, the trend and fluctuation of past data are modeled and future DLR values are estimated. Autoregressive model and its variants are expanded to reach accurate forecasting. Such methods apply white noise assumption to account for the DLR fluctuations. Since DLR fluctuations are related to weather condition, the white noise assumption cannot model fluctuations correctly. The Brownian motion has been implemented to meet data fluctuation issues in time series prediction. The Ornstein–Uhlenbeck (OU) process is one of the most widely used group of forecasting methods which consider Brownian motion. However, this approach is able to model a single factor that has never driven over the time. Therefore, implementing this factor is not suitable for forecasting DLR. In this paper, the OU process is extended into an integrated factorized OU to model and predict DLR values by considering hidden factors of DLR such as the weather conditions. The results, which are evaluated by the reference models, illustrate significant improvement in performance of the points and fluctuations of DLR.