Wind Power Forecasting Model Using Deep Learning Approach

Wind power generation highly depends on weather conditions because of the variation of wind speed from time to time. There is no accurate and reliable forecasting model for the Adama wind farm found in Ethiopia, that enables to predict the power generated at each instant of time. The primary goal of this research is to develop a model for forecasting wind power for the Adama wind farm by using deep learning techniques. In order to achieve the highest forecasting accuracy, energy forecasting model should take appropriate historical data to show future trends. We have collected a total of four years data starting from 2016–2019 with 163,802 rows with five-minute interval. We have used powerful Deep Learning approaches such as Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BiLSTM) for developing a wind power forecasting model. The Experimental result of LSTM model scores Mean Absolute Error and Mean Absolute Percentage Error of 0.644 and 0.340, respectively. Whereas, BiLSTM model scores Mean Absolute Error and Mean Absolute Percentage Error 0.643 and 0.327 respectively. This suggests that proposed deep learning models could predict the power from wind with reasonable accuracy.