A Probabilistic Estimation of Transmission Congestion for Mitigating Wind Power Curtailments

Abstract

The high penetration of renewable energy sources presents new challenges to power systems owing to the variability and uncertainty of renewable energy. When transmission congestion occurs, wind curtailment is required to ensure stability. Short-term wind power forecasting and transmission expansion are methods for managing the curtailment caused by transmission congestion. In addition, for transmission expansion, it is necessary to identify transmission congestion by assuming wind power scenarios. In this study, a short-term wind power forecasting model based on rolling long short-term memory (R-LSTM) is proposed for a 24-hour seasonal transmission congestion estimation. R-LSTM uses a recursive strategy to improve forecasting accuracy. Moreover, a method for probabilistic estimation of transmission congestion using seasonal hourly scenarios is proposed to account for variability and uncertainty. Wind power and electricity demand scenarios were generated using kernel density estimation (KDE) and the Metropolis-Hastings algorithm, a type of Markov chain Monte Carlo (MCMC) sampling. The R-LSTM model and transmission congestion estimation use empirical data from wind farm A and electricity demand on Jeju Island, where renewable energy is rapidly increasing. The results show that R-LSTM has a higher accuracy than long short-term memory (LSTM). The probability estimation results of transmission congestion revealed the season and time period with the highest probability of transmission congestion occurrence for each congestion components, as well as the lowest seasons and time periods.

Keywords

• Identifying transmission congestion
• kernel density estimation
• long short-term memory
• Metropolis-Hastings algorithm
• probabilistic estimation