Short Term Photovoltaic Power Combined Forecasting Based on Feature Extraction

The proposed method aims to address the randomness and volatility of photovoltaic power generation by forecasting short-term photovoltaic power combinations using feature extraction. The method incorporates the firefly sparrow algorithm to optimize the parameters of variational mode decomposition and obtain the signal component with the best forecasting effect. Because the signal components contain different influence characteristics and time series information, the maximum information coefficient is used to screen the characteristics of each signal component, and the feature matrix is established. The zero-crossing rate is used as the index to determine the demarcation point of high and low-frequency signals. The proposed model in the paper aims to enhance the accuracy and stability of short-term photovoltaic power generation forecasting by leveraging the distinct characteristics of high and low-frequency signals. It achieves this by combining the strengths of the long short-term memory neural network and the multi-layer perceptron to handle high and low-frequency signal components, respectively. The model's performance is compared with a baseline model called VMD-LSTM-MLP, as well as other existing forecasting models, under various weather conditions.