Volume 59

Long-term photovoltaic power generation prediction based on GNN-LightTS Ke Yan, Jiazhen Zhang, Kangning Wu, Qianyu Yang, Ying Li

Abstract

Accurate prediction of photovoltaic power generation is crucial for the effective integration of renewable energy into the power grid, which is beneficial for real-time balance of the power grid and optimization of energy storage systems. However, due to the intermittency and volatility of photovoltaic power generation, achieving accurate photovoltaic prediction remains a challenge, especially for long-term photovoltaic prediction. This paper proposes a new method for long-term photovoltaic power generation prediction. This method innovatively combines graph neural network and LightTS model. Experiments were conducted using the proposed model on datasets from three photovoltaic power stations characterized by distinct terrain and climate conditions. The results show that our model surpasses all benchmark models in long-term forecasting and also achieves notable superiority in medium- and short-term predictions.

Keywords Long-term photovoltaic power generation prediction, GNN, LightTS