A Comparative Analysis of Neural Network Approaches in Renewable Energy Prediction

The worldwide growing energy consumption is driving an essential transition to renewable energy alternatives like solar and wind power. However, a precise integration of these sources is crucial due to inherent variability, affecting the cost of optimal integration into power grid. This study presents a comparative analysis of statistical and deep learning techniques for renewable energy forecasting, focusing on solar irradiance and wind speed. Traditional statistical models such as ARIMA and SARIMA are evaluated alongside advanced deep learning models, including ESN, LSTM, bi-LSTM, Encoder-Decoder LSTM, CNN, CNN-LSTM, and LSTM-CNN. The implemented models are trained and tested using data from multiple locations in India. Results indicate that deep learning models outperform statistical methods, with CNN-based hybrid models achieving the best accuracy for wind speed forecasting and LSTM variants demonstrating superior performance in GHI prediction. The outcome of this study suggests neural network models at identifying complex time-based relationships and nonlinear trends in energy data.