山东电力技术2025，Vol.52Issue(12)：1-16,16.DOI:10.20097/j.cnki.issn1007-9904.240364
基于偏最小二乘和改进时间卷积网络的风电场超短期发电功率预测
Ultra-short Term Power Generation Prediction of Wind Farm Based on Partial Least Squares and Improved Time Convolutional Network
摘要
Abstract
To enhance the accuracy of ultra-short-term wind power prediction,a method combines the partial least squares(PLS)technique with an improved temporal convolutional network(ITCN)is proposed.First,PLS is employed to extract meteorological features that are closely related to wind power.Next,a spatial convolution layer is integrated into the temporal convolutional network(TCN)to improve its ability to extract both temporal and spatial features.To prevent local optimization,the slime mold algorithm is utilized for hyperparameter optimization.Experimental results from actual datasets in eastern and western China demonstrate that the proposed method effectively captures input variables and spatiotemporal characteristics of wind power,improving prediction accuracy by 12.61%,4.45%,3.65%,and 2.18%compared to common methods(CNN,LSTM,BiLSTM,and TCN).This method offers a novel approach to enhancing the efficiency and stability of wind farm operations.关键词
风电功率预测/偏最小二乘法/改进的时间卷积网络/空间卷积层/黏菌算法优化Key words
wind power prediction/partial least square method/improved time convolutional network/spatial convolution layer/slime mold algorithm optimization分类
能源科技引用本文复制引用
ZHENG Guangyu,XIN Zheng,CHI Weiran,CHEN Lizheng,WANG Shibo..基于偏最小二乘和改进时间卷积网络的风电场超短期发电功率预测[J].山东电力技术,2025,52(12):1-16,16.基金项目
山东省自然科学基金青年科学基金项目"考虑极端气象的智能建筑微电网与电力系统动态交互特性研究"(ZR2021QF066).Shandong Provincial Natural Science Fund Youth Science Fund Project"Research on Dynamic Interaction Characteristics between Intelligent Building Microgrids and Power Systems Considering Extreme Weather"(ZR2021QF066). （ZR2021QF066）
