A multi-timescale photovoltaic power prediction method based on SE-CNN-BiLSTM and improved Transformer
摘要
The random and volatile distributed photovoltaic （PV） power pose challenges for accurate forecasting and dispatch decision-making in power system operations. To address this， A multi-timescale photovoltaic power prediction method based on SE-CNN-BiLSTM and improved Transformer is proposed. Firstly， leveraging the diurnal trend similarity characteristics of PV power， a feature extraction method incorporating a channel attention mechanism is proposed to construct a prediction model for PV power trend features. Subsequently， based on the short-term fluctuation characteristics of PV power， a fluctuation feature extraction method based on similar time-period matching （STM） is proposed， utilizing the weather-induced fluctuation features of PV power to build a prediction model based on an improved Transformer. Then， by fusing the long- and short-timescale trend features and fluctuation features of PV power， a multi-timescale fusion method for PV power prediction is constructed. Finally， the proposed model is validated using actual operational data from a PV power station and simulation data. Results demonstrate that the proposed method effectively enhances the representational capacity and prediction accuracy of the forecasting model.