Triple Generative Adversarial Networks Based Weather Classification Model for Short-Term Photovoltaic Power Forecasting

Abstract

The challenge in the photovoltaic power (PV) forecasting is insufficiency of measuring dataset, especially for the extreme weather conditions. To alleviate this issue, Triple generative adversarial networks (Triple-GAN) based weather classification model is presented to predict short-term PV solar energy in a semisupervised manner. The simulation results demonstrate that Triple-GAN can improve diversity of the data and capture the intrinsic features of the measuring weather data. Meanwhile, the accuracy of weather classification is improved and the weather classification model is robust for short-term PV forecasting