A multi-stage forecasting system for daily ocean tidal energy based on secondary decomposition, optimized gate recurrent unit and error correction

Abstract

Tidal energy, as a new energy, has very high research potential and practical application value. For the characteristics of tidal energy such as nonstationarity and nonlinearity, a multi-stage forecasting system for daily ocean tidal energy is proposed. It is based on improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), refined composite multi-scale dispersion entropy (RCMDE), empirical mode decomposition based on time-varying filter modified by white shark optimizer (WSOTVFEMD), improved gate recurrent unit using parasitic salp swarm algorithm based on differential evolution (PDESSAGRU) and error correction using CNN (CNN-EC), named as ICEEMDAN-RCMDE-WSOTVFEMD-PDESSAGRU–CNN–EC. Firstly, decompose tidal energy into a series of intrinsic mode functions (IMFs) by ICEEMDAN, and divide IMFs into high-complexity and low-complexity components by RCMDE. Next, secondly decompose the reconstructed high-complexity components into high-complexity parts by WSOTVFEMD. Afterwards, separately predict each component of the high-complexity parts and the low-complexity components by PDESSA, and reconstruct the predicted results to obtain original predicting results. In the end, decompose the error into error IMFs (EIMFs) by ICEEMDAN, predict EIMFs with convolutional neural network (CNN) respectively to acquire error predicting results, and reconstruct original predicting results and error predicting results to acquire the final results. Taking the tidal energy of the American cities including San Francisco, Sitka, and Wauna as case studies, the results show that the proposed system has high prediction accuracy after experiments with 13 comparative models in each city.