A Photovoltaic Power Forecasting Model Based on VMD-ISSA-Transformer-GRU

To address the volatility and non-linear characteristics of photovoltaic power data, a photovoltaic power forecasting model is proposed. This model employs a hybrid architecture that optimises the hyperparameters of a Transformer-GRU fusion mechanism through Variational Mode Decomposition (VMD) and an Improved Sparrow Search Algorithm (ISSA). Firstly, historical photovoltaic data undergoes effective feature decomposition via VMD. Secondly, the ISSA optimises hyperparameters of the Transformer-GRU model, including learning rate, training cycles, and hidden layer size. Finally, the optimised Transformer-GRU generates photovoltaic power forecasts. Experimental results demonstrate that the VMD-ISSA-Transformer-GRU model achieves significantly higher prediction accuracy than GRU,Transformer-GRU ,VMD- Transformer-GRU and VMD-SSA-Transformer-GRU models.