考虑PCA-LSTM的风电机组输出功率预测研究  

Study on Prediction of Wind Turbine Output Power Considering PCA-LSTM

作　　者：李明[1] 袁逸萍[1] 贾依达尔 赵琴 LI Ming;YUAN Yi-ping;JIA Yi-daer;ZHAO Qin(School of Mechanical Engineering,Xinjiang University,Xinjiang Urumqi 830047,China)

机构地区：[1]新疆大学机械工程学院,新疆乌鲁木齐830047

出　　处：《机械设计与制造》2022年第9期145-148,共4页Machinery Design & Manufacture

基　　金：国家自然科学基金(51365054)。

摘　　要：为在风电机组多源异构SCADA大数据中挖掘有效信息,保证短期功率预测精度,针对数据在时间维度上的记忆特征,提出主成分分析(PCA)与长短期记忆神经网络(Long Short-Term Memory Neural Network,LSTM)相结合的短期功率预测模型。运用Pearson相关系数将海量的负荷数据、气象数据等相关数据进行初步提取,运用PCA技术降低数据维度,再将压缩后的数据输入LSTM网络进行短期负荷预测。实验结果表明,所提的方法比经典的时序预测方法如Arima、SVM等具有更高的精度,具有更稳定的预测结果。In order to mine effective information in multi-source heterogeneous SCADA big data of wind turbines to ensure the ac⁃curacy of short-term power prediction,a short-term power prediction model combining principal component analysis and longterm and short-term memory neural networks was proposed for the memory characteristics of the data in the time dimension.The pearson correlation coefficient is used to extract the massive load data,meteorological data and other related data.The PCA tech⁃nology is used to reduce the data dimension,and the compressed data is input to the LSTM network for short-term load predic⁃tion.Experimental results show that the proposed method has higher accuracy and more stable prediction results than the classic time-series prediction methods such as Arima and SVM.

关 键 词：风电机组 SCADA大数据 主成分分析 长短期记忆神经网络 预测 

分 类 号：TH16[机械工程—机械制造及自动化] TH460.5040