SPLTRAK Abstract Submission
Study of ANN Input Variables to Improve Hourly Day-ahead Solar Energy Prediction for Supporting Smart Grid Implementation in Semau Island
Ignatius Rendroyoko1,3, Hugo Hadi Suhana2,3, Ngapuli Irmea Sinisuka1, Eddie Widono Suwondho4
1Bandung Institute of Technology, Bandung, Indonesia
/2Trisakti University, Jakarta, Indonesia
/3PT Icon+, Jakarta, Indonesia
/4Prakarsa Jaringan Cerdas , Jakarta, Indonesia

Abstract—Solar power prediction is a crucial  aspect for power generation planning as well as supporting operations in a grid area that applies photovoltaic (PV) generators. Artificial neural networks (ANN) is known as a powerful method used for that task. In which, its performance is affected by satisfactorily of considered input variables. This paper studies input variables of ANN to result a better performance. The new input variables proposed are corresponding to both daily meteorological and air condition data, e.g. rainfall and air situations, respectively. For simulation purposes, irradiance prediction with difference input variables are carried out. Exploring the performance of the model for the best and worst case scenarios reveal that on both input variables are achievable with strong positive linear correlations.  Finally, both the proper variable inputs selected and the proposed method show good performance so as achieving the above mentioned prediction purposes.  
Keywords—artificial neural network, input variables, operation, planning, prediction.