Frontiers in Energy Research (Jun 2023)

Optimized ensemble model for wind power forecasting using hybrid whale and dipper-throated optimization algorithms

  • Amel Ali Alhussan,
  • Alaa Kadhim Farhan,
  • Abdelaziz A. Abdelhamid,
  • Abdelaziz A. Abdelhamid,
  • El-Sayed M. El-Kenawy,
  • Abdelhameed Ibrahim,
  • Doaa Sami Khafaga

DOI
https://doi.org/10.3389/fenrg.2023.1174910
Journal volume & issue
Vol. 11

Abstract

Read online

Introduction: Power generated by the wind is a viable renewable energy option. Forecasting wind power generation is particularly important for easing supply and demand imbalances in the smart grid. However, the biggest challenge with wind power is that it is unpredictable due to its intermittent and sporadic nature. The purpose of this research is to propose a reliable ensemble model that can predict future wind power generation.Methods: The proposed ensemble model comprises three reliable regression models: long short-term memory (LSTM), gated recurrent unit (GRU), and bidirectional LSTM models. To boost the performance of the proposed ensemble model, the outputs of each model are optimally weighted to form the final prediction output. The ensemble models’ weights are optimized in terms of a newly developed optimization algorithm based on the whale optimization algorithm and the dipper-throated optimization algorithm. On the other hand, the proposed optimization algorithm is converted to binary to be used in feature selection to boost the prediction results further. The proposed optimized ensemble model is tested in terms of a dataset publicly available on Kaggle.Results and discussion: The results of the proposed model are compared to the other six optimization algorithms to prove the superiority of the proposed optimization algorithm. In addition, statistical tests are performed to highlight the proposed approach’s performance and effectiveness in predicting future wind power values. The results are evaluated using a set of criteria such as root mean square error (RMSE), mean absolute error (MAE), and R2. The proposed approach could achieve the following results: RMSE = 0.0022, MAE = 0.0003, and R2 = 0.9999, which outperform those results achieved by other methods.

Keywords