Artificial Neural Network Performance Analysis for Solar Radiation Prediction, Case Study at Baron Techno Park

Abstract

As stated in Government Regulation No. 79 of 2014 on National Energy Policy (KEN), the New and Renewable Energy (NRE) mix target is at least 23% by 2025. Now the utilization of solar energy in Indonesia has only reached about 0.05% or 100 MW. Government compiles a roadmap for the use of solar energy that targets the installed PV mini-grid capacity until 2025 is 6.5 GW. Technically, before installing a solar power plant, solar potential data is needed for a certain period of time. This is absolutely necessary considering the potential for solar is intermittent. The solar data are then processed to create a model forecasting so that it can optimize the resulting energy output. Forecasting using artificial intelligence with artificial neural network algorithms is a good solution because it has higher accuracy. To see the comparison of the performance of the ANN of this research with previous research. The finding shows that Baron 1-7-1 performed better than 2P 1-2-1 and Baron 1-2-1, regardless that the RMSE have a slight difference but still Baron 1-7-1 outperformed the others, with the best value of RMSE 0.15185 and R2 of 0.88996.