Time-Series Model for Forecasting Short-term Future Additions of Renewable Energy to Worldwide Capacity

In this paper, we present an estimation model for forecasting the future net capacity additions of renewable energy based on the past time series data from 2000 to 2019. We use the regression based shallow neural network (R_SNN) model with 20 hidden neurons to estimate the best possible training parameters (weights and biases) for the model that provide the highest accuracy in model fitting and time series data prediction. R_SNN (20) model has been successfully used to model the time series data and forecast short-term additions of renewable energy in the next six years from 2020 to 2025. Results show that R_SNN (20) model has recorded the minimum loss based MSE (mean square error) for model estimation only after 50 epochs of neural network training with highest data fitting accuracy of 99%. The model shows a linearly increasing trend in the addition of renewable energy capacity in the following years from the year 2020-2025.