Large-Scale Solar Energy Generation Forecasting using Graph Neural Network

Solar energy is radiant energy produced by the sun, which radiates and emits enormous amounts of energy every day. Many technologies are used today to harness the energy of the sun, including photovoltaic system. It has promising potential for building large-scale solar power plants. Solar energy forecasts have a significant impact on decisions related to the operation and management of energy systems. Accurately predicting renewable energy performance is critical in ensuring grid reliability, resilience, reducing energy market and system risks. Deep learning (DL) recent success in many applications has attracted researchers implementing DL in solar energy forecasting tasks. Graph Neural Network (GNN) is one of the DL methods that perform inference on data described by graphs. The proposed work implemented GNN method to forecast the solar energy generation of LargeScale Solar (LSS) farm specifically for LSS Sepang plant. In GNN, the variables in the dataset were described by a self-learning graph structure. Four performance metrics were used to assess the model performance, such as root mean squared error (RMSE), mean squared error (MSE), mean absolute error (MAE) and accuracy. GNN model has demonstrated an effective solar energy generation forecasting model with the lowest value of MAE (537.28), MSE (0.07), RMSE (0.26) and highest accuracy (93.42%), that shows the great potential of the GNN method in forecasting the solar energy generation of LSS farm.