电力系统及其自动化学报2025，Vol.37Issue(10)：64-74,11.DOI:10.19635/j.cnki.csu-epsa.001588
时间生成对抗网络样本扩增下的风电功率预测
Wind Power Prediction Based on Sample Amplification of Time-series Generative Adversarial Network
摘要
Abstract
Due to the failure and aging of equipment in wind farms,it becomes difficult to obtain high-quality data sam-ples,which affects the accuracy of wind power prediction.To improve the accuracy of wind power prediction,the Laida rule was used for data cleaning and preprocessing at first,and then the dataset was divided.Second,the Pearson corre-lation coefficient method was used to screen the features,retain the strong features and eliminate the redundant fea-tures.Third,the time-series generative adversarial network was used to amplify the data and generate more samples.Fi-nally,a prediction model was constructed by combining the bidirectional temporal convolutional network,bidirectional gated recurrent unit and attention mechanism.Experiments were carried out on the power and meteorological data of one wind farm in Hami,Xinjiang,and it was concluded that the proposed model can effectively improve the accuracy of wind power prediction and solve the problems of insufficient sample data and outlier interference.关键词
风电功率预测/皮尔逊相关系数/生成对抗网络/时域卷积网络/门控循环单元Key words
wind power prediction/Pearson correlation coefficient/generative adversarial network/temporal convolu-tional network/gated recurrent unit分类
信息技术与安全科学引用本文复制引用
张智浩,王聪,张宏立,马萍,李新凯..时间生成对抗网络样本扩增下的风电功率预测[J].电力系统及其自动化学报,2025,37(10):64-74,11.基金项目
新疆维吾尔自治区自然科学基金资助项目(2023D01C187).天山英才培养计划(2023TSYCCX0037). （2023D01C187）
