江苏省科技期刊可视化阅读平台

导航条

《无线互联科技》杂志社 ›› 2025, Vol. 22 ›› Issue (1): 63-66.

• 技术应用 • 上一篇    下一篇

LSTM与XGBoost混合模型在风力发电功率预测中的应用

陈大为1, 张玮1, 慕龙2   

  1. 1.武威新诚新能源有限公司,甘肃 武威 737100;
    2.古浪绿舟光伏发电有限公司,甘肃 武威 737100
  • 出版日期:2025-01-10 发布日期:2025-03-21
  • 作者简介:陈大为(1983— ),男,工程师,本科;研究方向:新能源发电项目建设及运行管理。

Application of LSTM and XGBoost hybrid model in wind power forecasting

CHEN Dawei 1, ZHANG Wei 1, MU Long2   

  1. 1. Wuwei Xincheng New Energy Co., Ltd., Wuwei 737100, China;
    2. Gulang Lvzhou Photovoltaic Power Generation Co., Ltd., Wuwei 737100, China
  • Online:2025-01-10 Published:2025-03-21

PDF

227

摘要： 文章提出了一种基于LSTM与XGBoost的混合模型用于风力发电功率预测。主要研究了LSTM模型与XGBoost模型的融合方法,通过LSTM捕捉序列数据的长期依赖关系,再利用XGBoost进行非线性拟合以提升预测精度。实验采用国家电网新能源发电预测大赛提供的公开数据集,使用平均绝对误差和决定系数等指标对模型性能进行评估。实验结果表明,文章所提出的混合模型相比标准LSTM模型在预测精度和拟合能力上均表现出显著的优势。

关键词: 长短期记忆, 极端梯度提升, 风力发电, 功率预测

Abstract: The article proposes a hybrid model for wind power forecasting based on LSTM and XGBoost, and investigates the fusion method of LSTM and XGBoost models. The LSTM captures the long-term dependencies of sequential data, while XGBoost is utilized for nonlinear fitting to enhance prediction accuracy. The experiment utilizes the public dataset provided by the State Grid New Energy Power Generation Forecasting Competition, and evaluates the model performance using metrics such as mean absolute error and determination coefficient. The experimental results demonstrate that the proposed hybrid model exhibits significant advantages over the standard LSTM model in terms of prediction accuracy and fitting capability.

Key words: long and short-term memory, extreme gradient boosting, wind power generation, power prediction

中图分类号: