摘要
本文利用自适应噪声的完备经验模态分解(CEEMDAN)对风电原始序列信号进行处理后,采用GRU-XGBoost模型对非线性、非平稳的功率序列进行建模和预测,以提升模型的预测能力和泛化性。首先,通过CEEMDAN将风电功率原始序列分解为不同时间尺度的分量,然后将分解后的信号输入GRU神经网络生成预测信号,最后通过XGBoost进行校正。通过与多种预测模型进行比较,证明了该模型在预测精度方面的卓越表现。
In this paper,after processing the original sequence signal of wind power using the adaptive noise complete Empirical Mode decomposition(CEEMDAN),the GRU-XGBoost model is used to model and predict the nonlinear and non-stationary power series,so as to improve the prediction ability and generalization of the model.First,the original sequence of wind power is decomposed into components of different time scales by CEEMDAN.The decomposed signal is then fed into the GRU neural network to generate a prediction signal.Finally,it is corrected by XGBoost.The excellent performance of this model in forecasting accuracy is proved by comparing with many forecasting models.
作者
耿运涛
Geng Yuntao(Shaoyang Polytechnic,Shaoyang 422000,Hunan,China)
出处
《船电技术》
2024年第7期32-35,共4页
Marine Electric & Electronic Engineering
基金
2023年湖南省教育厅科学研究项目《高比例风电渗透率电力系统频率特性研究》(项目编号:23C0954)。
