A Short-term Wind Power Output Forecasting Model based on the Enhanced Gradient Boosting Machine (GBM) Algorithms for High Wind Power Penetrations

With rising concerns of climate change, there has been a worldwide trend of establishing policies regarding net zero emissions and sustainability. According to Korea's 2050 Carbon Neutral Strategy, the government aims to decarbonize the country's economic structure and increase penetration of renewable energies. Statistics also show that wind power generation in Korea has been increasing steadily over the years. However, the intermittent nature of wind power remains an obstacle in predicting wind power outputs. Therefore, accuracy in wind power forecasts must be improved to facilitate larger integration of renewables to existing electrical grids. In this paper, we propose the implementation of a short-term wind power output forecasting model based on the enhanced Gradient Boosting Machine (G B M) algorithms for high wind power penetrations. G B M is an effective machine learning algorithm which improves its performance by combining previously learned weak learners to form a strong learner. A 15-minute cycle of measured data from Jeju's wind farms is applied to the model as the input data. The results include scatter plots and line graphs depicting the outcome of prediction data by the G B M model and real data.