Research on short-term power forecasting method of photovoltaic generation based on BWO-optimized VMD-KELM

With the intensification of global energy supply tensions and environmental pollution issues, solar energy, as a green and sustainable energy option, is gaining increasing attention. However, the output power of photovoltaic (PV) generation is susceptible to various factors, exhibiting randomness, volatility, and uncertainty, which poses significant challenges for accurate forecasting. To enhance the accuracy of predictions, this paper proposes a model that combines the Beluga Whale Optimization (BWO) algorithm with Variational Mode Decomposition (VMD) and the Kernel Extreme Learning Machine (KELM). The model first employs VMD technology for preprocessing photovoltaic power generation data, aiming to reduce the non-stationarity of the data; subsequently, the BWO algorithm is utilized to fine-tune the parameters of the KELM model to enhance its predictive performance. Finally, the model was trained and validated using actual photovoltaic power generation data. The findings reveal that the proposed BWO-VMD-KELM model demonstrates superior performance over conventional KELM and VMD-KELM models. Specifically, on the training set, the model achieves Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (R^2) values of 16.316, 23.424, and 0.963, respectively. These results are indicative of the model's enhanced predictive accuracy and its overall effectiveness and superiority in the context of photovoltaic power forecasting.