Solar Irradiance Prediction with Multi-Meteorological Variables Based on iTransformer

Photovoltaic power generation continues to grow globally and plays a crucial role in energy transformation and sustainable development. By accurately predicting irradiance, photovoltaic power generation systems can better adjust their own working conditions, improve power generation efficiency, reduce energy production costs, and better meet energy demand and promote the use of clean energy. This article selects Informer and Pyraformer models that improve the traditional attention mechanism based on Transformer, as well as iTransformer models that redesign the Transformer architecture without modifying any Transformer modules, to study their effectiveness in multivariate long-term irradiance prediction. The results indicate that the iTransformer model has achieved more advanced performance in multivariate long-term irradiance prediction problems, which provides a good case for subsequent research and a promising direction for improving Transformer based predictors.