Ultra‐Short‐Term Power Forecasting of Single‐Axis Tracking PV System Based on Irradiance Transformation and CNN‐BiGRU Model

ABSTRACT PV power forecasting is essential for the stable and efficient operation of power systems. However, forecasting accuracy is compromised by two main factors: the quality of solar irradiance predictions and the limited flexibility of forecasting methods. In this paper, an ultra‐short‐term PV power forecasting method for the SAT PV system is proposed based on irradiance transformation and the CNN‐BiGRU model. First, a SAT‐based irradiance transformation is adopted to convert the horizontal‐plane NWP irradiance data into the effective irradiance received by the SAT PV panels. Then, the FCM clustering algorithm is used to classify weather types, improving the adaptability of forecasting models to diverse meteorological conditions. Finally, a hybrid CNN‐BiGRU network is developed to capture spatiotemporal features for accurate power prediction. The proposed method is validated using two‐year operational data from a 50 MW PV plant in Northwest China and a 60 MW plant in Southeast China. A high degree of forecasting accuracy is observed. For the Northwest plant, the MSE values are 2.46, 4.53 and 3.64 MW 2 under sunny, cloudy and rainy conditions, respectively; corresponding results for the Southeast plant are 2.62, 4.84 and 3.80 MW 2 . Consistent improvements are also observed in MAE, RMSE, R 2 and FS metrics. Comparative tests confirm the superiority of the proposed model over state‐of‐the‐art models, demonstrating that the integration of irradiance transformation and CNN‐BiGRU model significantly enhances forecasting performance for SAT PV systems.