Performance of Stacked LSTM for Solar Energy Revenue Prediction

This study investigates the performance of stacked and simplified Long Short-Term Memory (LSTM) models for predicting daily revenue from a 1 MW solar power plant at Pandit Deendayal Energy University (PDEU), Gujarat, India. Utilizing the meteorological and operational data, the models were evaluated using root mean squared error (RMSE) and mean absolute error (MAE). The stacked LSTM demonstrated enhanced accuracy, capturing complex temporal dependencies more effectively than the simplified model. The findings highlight the potential of stacked LSTM for improving financial planning and operational efficiency in forecasting different attributes of solar energy generation.