Research on medium-term electric power load forecasting based on SSAD-CEEMDAN

Abstract

To fully enhance the accuracy of electric power load forecasting, a medium-term electric power load forecasting model based on secondary decomposition (predicted model of power load based on secondary decomposition, PMPL-SD) is proposed. In the PMPL-SD algorithm, the original load data are first decomposed and reconstructed using the singular spectrum analysis decomposition method; then, the data after denoising are further decomposed and reconstructed using the noise-assisted complete ensemble empirical mode decomposition to obtain three components: high frequency, medium frequency, and trend. At the same time, a combined network model consisting of convolutional neural networks and bidirectional long short-term memory networks is used to predict the three components separately. The component results are integrated to obtain the final forecasting result. To verify the performance of the PMPL-SD algorithm, three models are selected for comparison. The experimental results show that the proposed algorithm has higher forecasting accuracy in medium-term electric power load forecasting.