Day-ahead hourly photovoltaic generation forecasting using extreme learning machine

Abstract

The photovoltaic (PV) generation systems as environmentally friendly renewable energy sources are increasing. However, the power generation of solar has high uncertainty and intermittency and brings significant challenges to power system operators. The accurate forecasting of photovoltaic (PV) power production is good for both the grid and individual smart homes. In this paper, we propose a novel weather-based photovoltaic generation forecasting approach using extreme learning machine (ELM) for 1-day ahead hourly forecasting of PV power output. In the proposed approach, the weather conditions are divided into three types which are sunny day, cloudy day, and rainy day and training the PV power output forecasting models separately for those three weather types. In this paper, we take the PV output history data from the PV experiment system located in Shanghai for case study. The forecasting results show that the proposed model outperform the BP neural networks model in all three weather types.