Ultra-short-term Prediction of Wind Power based on QR-BLSTM

To address the problems that existing ultra-short-term prediction methods are difficult to effectively mine and analyze the output fluctuation characteristics of wind energy, this paper proposes a wind energy ultra-short-term conditional probability prediction method based on quantile regression-bid...

Full description

Saved in:
Bibliographic Details
Published in2023 8th Asia Conference on Power and Electrical Engineering (ACPEE) pp. 1385 - 1389
Main Authors Lu, Qiqi, Wang, Hao, Luo, Jiahao, Liu, Zhishuai, Wang, Shuaiwei
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.04.2023
Subjects
bidirectional long short-term memory
Data models
Deep learning
nonlinear quantile regression
Predictive models
Probabilistic logic
probabilistic prediction
Time series analysis
Wind energy
wind power
Wind power generation
Online AccessGet full text

Cover

More Information
Summary:To address the problems that existing ultra-short-term prediction methods are difficult to effectively mine and analyze the output fluctuation characteristics of wind energy, this paper proposes a wind energy ultra-short-term conditional probability prediction method based on quantile regression-bidirectional long short-term memory (QR-BLSTM). Firstly, conduct correlation analysis and feature dimensionality reduction on raw power data through correlation analysis, and calculate data under each quantile based on quantile regression. Then, establish a BLSTM model for point prediction by combining two sets of LSTMs in an opposite form. Next, analyze and optimize the distribution of point prediction errors, and calculate the probability intervals under each prediction point, taking into account the conditional quantile results. Finally, validate the model based on the measured data of the wind farm. The results show that the proposed method has a good prediction performance with an accuracy of over 97% in the 2-h ahead prediction.
DOI:10.1109/ACPEE56931.2023.10135773