FORECASTING SHORT-TERM POWER OUTPUT OF PHOTOVOLTAIC SYSTEMS UNDER EQUATORIAL CLIMATE CONDITIONS

摘要

Accurate and reliable forecasting of photovoltaic system's power output is an essential supporting technique that will facilitate grid-connected photovoltaic systems. This paper aims to evaluate several artificial-intelligence-based forecasting techniques to predict the power production of a 24 kW photovoltaic system in Melaka, Malaysia. This location is influenced by equatorial climate conditions, having no distinct seasonal variation and high amount of cloud cover. The data comprises five-minute averaged values of global irradiance. backmodule temperature, and inverter AC output, recorded throughout the year. Findings show that LS-SVM produced the best forecast in terms of the evaluation metrics used followed by the original SVM at a close second.