Recurrent Neural Network-Based Solar Power Generation Forecasting Model in Comparison with ANN

The usage of renewable energy has increased recently which makes power sector company plan ahead for the long term. Hence, forecasting is necessary but all renewable energy system is nature dependent which introduces non-linearity in the system. To forecast nonlinear data machine learning and computation techniques are used. In this work, a memory-based recurrent neural network is developed and compared with a traditional ML model, that is, Artificial Neural Network (ANN). The models are forecasted for a month ahead and two months ahead. The performance of the model is evaluated over statistical parameters such as Mean Absolute Error (MAE), Mean Square Error (MSE), Mean Absolute Percentage Error (MAPE), and correlation of determination (R2) is used. The percentage improvement calculated for the RNN model over the ANN model is found to be 8.5412, 25.2926, 10.6301, and 5.8628 for a month ahead prediction, and for two months ahead prediction is found to be 12.5920, 26.3174, 32.3866 and 8.3180, considering statistical parameters data. This result strongly supports that RNN based prediction model is superior to ANN, and can be used for long-term ahead planning.