Short-Term Solar Power Forecasting Using SVR on Hybrid PV Power Plant in Indonesia

Abstract

Considering the environmental issues, the use of renewable energy sources is a far more sustainable solution to meeting the energy demand than fossil fuels. However, the limited availability of renewable energy is a growing problem to be solved. Solar energy has become a popular renewable energy source in several countries such as Indonesia because of their equatorial locations. In this study, limited meteorological measurement has been applied with the aim of forecasting solar power generation for planning photovoltaic (PV) power plants, especially in rural areas, which have limited access to fossil energy. We used limited measurements such as temperature, humidity, and solar radiation. The use of support vector regression (SVR) was applied to improve denoising capabilities and simplify computation. SVR has been evaluated using statistical metrics such as mean absolute percentage error (MAPE), relative root means square error (NRMSE), and coefficient of determination (R2). The results showed the MAPE value obtained 18.56% from the RBF_SVR. NRMSE value performed excellently with 8.02% from the SW-SVR method. R2 also indicated good forecasting with 0.99. The results showed that promising short-term solar power generation forecasting can be applied to estimate the availability of solar power, plan for an extension, and assess the performance of hybrid power plants in Indonesia.