Development of Short Term Solar Radiation Forecasting Using AI Techniques

Solar energy (SE) is one of the more popular renewable energy sources, but its integration into the grid is complicated due to its intermittent nature and challenges associated with its generation. South Africa, with an annual global horizontal irradiance of 200W/m2, has a great deal of potential for utilizing SE as a source of energy. Accurate SE forecasting is critical for large-scale SE production to grow. Artificial intelligence techniques may prove to be a feasible solution for better forecasting of SE given its intermittent nature and other challenges including weather, temperature, etc. This article presents the view of real-time data collection of various stations located in South Africa and to design the SE forecasting using an an adaptive neuro-fuzzy inference system (ANFIS) and an artificial neural network (ANN) and for forecasting global horizontal irradiance in the short term (GHI). The forecasting models are evaluated via mean absolute square error (MAPE), and root mean square error (RMSE). From the results, it is seen that the ANFIS model outperformed the ANN model for SE accurate forecasting with a lesser value of error. The outcomes of both models for SE are shown via graphical view and also through numerical data.