Short-Term Photovoltaic Power Combined Prediction Based on Feature Screening and Weight Optimization

Aiming at the problem of low prediction accuracy caused by the intermittent and fluctuating characteristics of photovoltaic power, a short-term photovoltaic power combined prediction method based on feature screening and weight optimization is proposed. Firstly, K-means is used to cluster the photovoltaic power; Secondly, CEEMDAN is used to decompose photovoltaic power and wavelet decomposition is used to decompose irradiance, and sample entropy and K-means are used to reconstruct each component of photovoltaic power into high, intermediate, and low frequency terms; Then, Spearman’s correlation coefficient is used to calculate the correlation between each meteorological factor and the decomposed irradiance component and the high, intermediate, and low frequency terms of photovoltaic power, and the feature selection is carried out; Then, CNN-BiLSTM-Attention is used to predict the high frequency term, LSTM is used to predict the intermediate frequency and low frequency terms, and the results are superimposed to obtain the preliminary prediction value; Finally, the dung beetle algorithm is used to optimize the weights of the initial prediction values of the training set of high, intermediate, and low frequency terms, and the optimal weight is substituted into the test set to obtain the final prediction result. The results show that the MAPE of the model is the smallest under different weather conditions, which can effectively improve the prediction accuracy.