Day-Ahead Solar Irradiance Forecasting Using Hybrid Recurrent Neural Network with Weather Classification for Power System Scheduling

At the present time, power-system planning and management is facing major challenge of integrating renewable energy resources (RESs) due to their intermittent nature. To address this problem, a highly accurate generation forecasting system needed for day-ahead power scheduling. Day-ahead solar irradiance (SI) has various applications operators market agents such as unit commitment, reserve management, biding in market. end, hybrid recurrent neural network presented herein that uses long short-term memory (LSTM-RNN) approach forecast SI. In approach, k-means clustering first used classify each day either sunny or cloudy. Then, LSTM-RNN learn uncertainty variability type cluster separately predict SI with better accuracy. The exogenous features dry-bulb temperature, dew point relative humidity are train models. Results show proposed model performed than feed-forward (FFNN), support vector machine (SVM), conventional LSTM-RNN, persistence model.