A Data-driven Method for Forecasting the Photovoltaic Substation Active Power

With the increasing need for accuracy in PV active power prediction, a new model combining variational mode decomposition (VMD), sparrow search algorithm (SSA), and long short-term memory neural network (LSTM) was established, named VMD-SSA-LSTM. Initially, VMD was utilized to break down historical PV active power and environmental sequence data Afterwards, LSTM parameters were optimized using SSA, and each decomposition component was input into the LSTM neural network. The forecasted outputs of each component were summed to derive the predicted PV active power value for the substation, which was verified with data from a specific region. Results indicate that, in comparison to the LSTM and VMD-LSTM models, the VMD-SSA-LSTM model achieves superior prediction accuracy, offering a new choice for predicting PV active power in substations.