Scholarlink

<p>Accurate wind speed forecast plays a vital role in smooth functioning and reliability of renewable energy platforms. Wind speed further helps in wind power production which enables the electrical system to efficiently manage its resources and ensure grid dependability. Besides, wind speed is intermittent and very unpredictable in nature, making reliable forecasting very difficult. In this study a hybrid framework integrating Variational Mode Decomposition (VMD) with a Gated Recurrent Unit (GRU) network for multi-step ahead prediction is developed to address above challenges. The VMD used to preprocess the data that decomposes the wind speed time series data into numerous sub modes. Further, GRU is used as predictor to forecast all sub-modes obtained from the VMD process and the estimated results of each sub-modes merged to provide final predictions. The study uses the grid search (GS) to optimize the hyperparameters of models instead of trial and error method. The developed models are evaluated on Jaisalmer and Muppandal wind datasets for one-, two-, and three-step ahead forecasting. For one-step forecasting, the VMD-GS-GRU obtained the lower RMSE of 0.228 and 0.084, MAE of 0.161 and 0.065, MAPE of 6.92% and 1.52%, and R<sup>2</sup> of 0.995 and 0.992 for Jaisalmer and Muppandal datasets respectively. The Diebold-Mariano (DM) test also confirms that the proposed model achieves statistically significant improvements over benchmark models at confidence levels ranging from 95 to 99%. Additionally, ramp event analysis demonstrates improved detection capability and reduced RMSE under dynamic conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multistep ahead wind forecast using variational mode decomposition and gated recurrent deep learning unit

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月7日

查看全文
  • Pardeep Singla,
  • Komal Duhan,
  • Sandeep Sharma,
  • Sumit Saroha,
  • Manoj Duhan,
  • Manoj Arora

摘要

Accurate wind speed forecast plays a vital role in smooth functioning and reliability of renewable energy platforms. Wind speed further helps in wind power production which enables the electrical system to efficiently manage its resources and ensure grid dependability. Besides, wind speed is intermittent and very unpredictable in nature, making reliable forecasting very difficult. In this study a hybrid framework integrating Variational Mode Decomposition (VMD) with a Gated Recurrent Unit (GRU) network for multi-step ahead prediction is developed to address above challenges. The VMD used to preprocess the data that decomposes the wind speed time series data into numerous sub modes. Further, GRU is used as predictor to forecast all sub-modes obtained from the VMD process and the estimated results of each sub-modes merged to provide final predictions. The study uses the grid search (GS) to optimize the hyperparameters of models instead of trial and error method. The developed models are evaluated on Jaisalmer and Muppandal wind datasets for one-, two-, and three-step ahead forecasting. For one-step forecasting, the VMD-GS-GRU obtained the lower RMSE of 0.228 and 0.084, MAE of 0.161 and 0.065, MAPE of 6.92% and 1.52%, and R2 of 0.995 and 0.992 for Jaisalmer and Muppandal datasets respectively. The Diebold-Mariano (DM) test also confirms that the proposed model achieves statistically significant improvements over benchmark models at confidence levels ranging from 95 to 99%. Additionally, ramp event analysis demonstrates improved detection capability and reduced RMSE under dynamic conditions.