Ning Zhou ^1* Bo-Wen Shang ^1* Jinshuai Zhang ¹ Shi-Hao Xu ^2* Ming-Ming Xu ^1*

• ¹ State Grid Henan Electric Power Research Institute, Zhengzhou, China
• ² School of Electrical Engineering, Shanghai University of Electric Power, Shanghai, China

Accurate prediction of photovoltaic power generation is of great significance to stable operation of power system. To improve the prediction accuracy of photovoltaic power, a photovoltaic power generation prediction machine learning model based on Transformer model is proposed in this paper. In this paper, the basic principle of Transformer model is introduced. Correlation analysis tools such as Pearson correlation coefficient and Spearman correlation coefficient are introduced to analyze the correlation between various factors and power generation in the photovoltaic power generation process. Then, the prediction results of traditional machine learning models and the Transformer model proposed in this paper were compared and analyzed for errors. The results show that: for long-term prediction tasks such as photovoltaic power generation prediction, Transformer model has higher prediction accuracy than traditional machine learning models. Moreover, compared with BP, LSTM and Bi-LSTM models, the Mean Square Error (MSE) of Transformer model decreases by 70.16%, 69.32% and 62.88% respectively in short-term prediction, and the Mean Square Error (MSE) of Transformer model decreases by 63.58%, 51.02% and 38.3% respectively in long-term prediction, which has good prediction effect. In addition, compared with the long-term prediction effect of Informer model, Transformer model has higher prediction accuracy.