Document Type : Original Article

Authors

• Ling Tan ¹
• Bin Wang ²
• Julong Chen ²
• Yongqing Zhu ²
• Junqiu Fan ²
• Jiang Hu ²

¹ organization={Nanjing University of Information Science and Technology}, addressline={219 Ningliu Road, Jiangbei New District}, city={Nanjing}, postcode={210044}, state={Jiangsu}, country={China}

² Power Grid Planning and Research Center Guizhou Power Grid Co, Ltd, 38 Ruijin South Road, Nanming District, Guiyang, 550003, Guizhou, China

Abstract

Medium- and long-term PV power forecasting is of great significance for the planning and management of new energy grids, and the existing medium- and long-term PV power forecasting methods generally suffer from the problems of insufficient processing means and low forecasting efficiency. Aiming at the challenges of weak spatial and temporal correlation of medium- and long-term PV power data, as well as data redundancy and low forecasting efficiency brought about by long-time forecasting, this paper proposes a medium- and long-term PV power forecasting method based on the Transformer, SP-Transformer (Spatiotemporal-ProbSparse Transformer), which aims to effectively capture the spatio-temporal correlation between meteorological and geographical elements and PV power. The method embeds the geographic location information of PV sites into the model through spatio-temporal location coding and designs a spatio-temporal probabilistic sparse self-attention mechanism, which reduces model complexity while allowing the model to better capture the spatio-temporal correlation between input data. To further enhance the model's ability to capture and generalize potential patterns in complex PV power data, this paper proposes a feature pyramid-based self-attention distillation module to ensure the accuracy and robustness of the model in long-term forecasting tasks. The SP-Transformer model performs well in the PV power forecasting task, with a medium-term (48 hours) forecasting accuracy of 93.8% and a long-term (336 hours) forecasting accuracy of 90.4%, both of which are better than all the comparative algorithms involved in the experiment.

Keywords

• PV power forecasting
• Medium and Long term forecasting
• Transformer
• Attention mechanism
• Feature pyramid self-attention distillation

Main Subjects

• Transformation of Generated Electricity by Renewable Energies to Grid