Photovoltaic power forecast model based on VMD-SSA-GRU

Improving the power estimation accuracy of photovoltaic power production grows proficient importance to ensure the power system's secure dispatch and sturdy operation. In the article, a photovoltaic power estimation approach is suggested by combining the variational mode decomposition (VMD), the sparrow search algorithm (SSA), and the gate recurrent unit (GRU). Based on completely regarding the four environmental attributions constraining the PV output power, firstly, VMD is used to disintegrate the sequence of environmental attributions to attain the data signals’ alterations in distinct time scales and decrease the non-stationarity of the sequence of environmental attributions; secondly, SSA is used to search for optimizing the GRU’s parameters; and finally, the GRU architecture is implemented to perform the dynamic temporal modeling on the sequence of multivariate features to achieve the PV power generation’s estimation. The final method is to utilize the GRU network for dynamic time modeling of the multivariate feature series to realize PV power’ prediction. The experimentation outcomes show that the prediction approach is more accurate than traditional PV power estimation methods.