Application of Machine Learning Forecasting Model for Renewable Generations of Adrar’s Power System

Uncertainties and intermittency of wind and solar power generations pose difficulties in power system operation. Thus, raises the importance of developing an accurate prediction model. This article proposes long short-term memory (LSTM) networks a complete tool to predict power output of two photovoltaic power plants and a wind farms installed in the isolated Adrar’s power system. Meteorological variables such as, solar irradiation, ambient and cellule temperatures, relative humidity, wind speed, and pressure, are introduced into LSTM for predicting power outputs. The LSTM is trained on the collected time series of real recorded dataset. The RMSE accuracy of LSTM networks forecasting methods is used to prediction performance of the proposed methods. The LSRTM can predict the power output of PV plants and wind turbine with a lowest RMSE of 12.48 kW is successively obtained for the wind turbine.