A multi-objective optimization-based ensemble neural network wind speed prediction model

doi:10.1016/j.ijepes.2025.110833

A multi-objective optimization-based ensemble neural network wind speed prediction model

2025-06-23

0

PRE

AI

DOI：10.1016/j.ijepes.2025.110833

Original

Original request for help

Save

Abstract

En 中文

• This paper uses a multi-variable prediction based on a sliding window, taking multiple feature variables from past moments as independent variables and future wind speed as the dependent variable. This approach gathers more information for prediction, offering greater accuracy than single-variable prediction. • This paper innovatively utilizes XGBoost to integrate neural network models and employs a novel multi-objective parameter optimization method to enhance the predictive accuracy of XGBoost. • This paper innovatively proposes a dynamically adaptive multi-objective optimization parameter algorithm, NS-ADPOA. This method addresses both the overfitting and underfitting issues of decision tree models. • The ensemble model in this paper effectively integrates various neural networks, including convolutional neural networks, recurrent neural networks, the latest ensemble models like Informer, and the original model proposed here: CNN-BiLSTM-AM. • This study validated the proposed model on four datasets representing diverse wind conditions, showing that it is capable of maintaining strong predictive performance across different wind environments.

Keywords:

Wind speed prediction

Neural network models

XGBoost algorithm

NS-ADPOA algorithm

Ensemble model

A multi-objective optimization-based ensemble neural network wind speed prediction model

Abstract

Journal

Organization