Short-Term Net Load Forecasting for Urban Distributed Photovoltaics Using SGMD and SE-Attention

To improve the accuracy of net load prediction after high proportion distributed photovoltaic grid connection, a photovoltaic substation net load prediction method based on symplectic geometric mode decomposition sample entropy and compressed excitation network is proposed. Firstly, a net load signal decomposition model based on fluctuation complexity partitioning was established. Through symplectic geometric mode decomposition and sample entropy partitioning, multiple stationary and complex symplectic geometric components (SGCs) with significant differences were obtained. Then, a screening model for SGC strong correlation prediction factors based on mutual information was established to determine the optimal prediction input for different SGCs. Finally, a combined prediction model for net load SGC based on compressed excitation network was established, targeting different SGCs and prediction factors, and combining the prediction results to achieve high-precision net load prediction. The analysis results based on actual data from a city in China show that the prediction error of the proposed method is 6.79<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, which is 2.64% lower than traditional methods.