Assessment of dynamic line rating forecasting methods

Abstract

• Forecasting electric power transmission line ampacity is presented here to be pivotal to achieving an increased line rating. • Deep learning techniques and regression were used to identify patterns in the environmental data and forecast DLR at different lead times. • This article compared forecasting techniques and evaluated them with point and probabilistic error metrics. • Point Error metrics gave the discrepancy between the forecasted and actual values, while the probabilistic error metrics gave the confidence level of the forecast values of the techniques. Optimal transmission line rating use is guaranteed with dynamic line rating (DLR). It is a smart grid technology that foresees variations in meteorological conditions affecting line rating and deploys algorithms to effect changes to the line rating due to these conditions. Electric power system operators use forecasted DLR for system planning, operation, and delivery. This study reviewed DLR forecasting techniques, classified them, implemented them, and compared their outputs at different lead times. It used ensemble means forecasting, recurrent neural network (RNN), and convolution neural network (CNN). Ensemble forecasting technique deployed in this study involves a Monte-Carlo simulation that produces random, equally viable predicting solutions. Alternatively, a neural network layer's initial outcome is fed back into it to predict the output in RNN, while CNN learns to predict features that vary in time and space with marginal discrepancies. This study used quantile regression (QR), ensemble forecasting, RNN and CNN to forecast DLR at 12hrs, 24hrs and 48hrs. The tested forecasting approaches prove efficient, but ensemble forecasting seems less error-prone, more secure and conservative among all methods. On average, 75th percentile quantile regression and ensemble forecasting demonstrate better reliability and avail us the better choice of ampacity among the forecasting techniques.