Predicting Transmission Line Congestion in Energy Systems with a High Share of Renewables

The increase of renewable electricity capacity and the intermittent nature of renewable generation create a constant mismatch between spatial generation and consumption patterns and the necessary transmission infrastructure. Therefore, congestion management strategies are becoming more and more vital for the electricity system. While markets in the United States or Norway traditionally implement market-based congestion management schemes, markets with a uniform market clearing price often rely on redispatch. Current congestion forecasts are computationally expensive and cannot be frequently updated as new weather information becomes available. We propose a forecasting mechanism based on an artificial neural network using only publicly available day-ahead data making a case against market based redispatch mechanisms. We validate the approach using empirical data and benchmark it against a Naïve classification method. We find that the algorithm performs well on the tested data predicting the majority of congested lines yielding high values of precision and recall.