计及动态时空相关性的多风电场短期功率预测  

Short-term Power Forecasting of Multiple Wind Farms Considering Dynamic Spatial-temporal Correlations

作　　者：李丹[1] 黄烽云 杨帆 唐建 罗娇娇 方泽仁 LI Dan;HUANG Fengyun;YANG Fan;TANG Jian;LUO Jiaojiao;FANG Zeren(College of Electrical Engineering and New Energy,China Three Gorges University,Yichang 443002,China;Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station,Yichang 443002,China;Yinchuan Power Supply Company,State Grid Ningxia Electric Power Co.,Ltd,Yinchuan 750000,China;Hubei Provincial Collaborative Innovation Center for New Energy Microgrid,Yichang 443002,China)

机构地区：[1]三峡大学电气与新能源学院,宜昌443002 [2]梯级水电站运行与控制湖北省重点实验室,宜昌443002 [3]国网宁夏电力有限公司银川供电公司,银川750011 [4]新能源微电网湖北省协同创新中心,宜昌443002

出　　处：《电力系统及其自动化学报》2025年第2期1-9,共9页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(51807109)。

摘　　要：针对同一区域内多风电场出力间复杂且动态的时空相关性,提出一种基于注意力时空同步图卷积网络的多风电场短期功率预测模型。首先引入注意力机制量化天气特征对风功率的影响,构建相邻3个时间步的风功率局部时空图,卷积提取局部时空特征;然后用时空同步图卷积层聚合输入时窗的整体时空特征;最后非线性映射输出多风电场未来时段的功率预测结果。实际算例结果表明,所提模型通过学习不同天气条件下风功率的时空动态演变规律,可将多风电场日前功率预测精度提高2.10%~13.94%。Aimedat the complex and dynamic spatial-temporal correlations between the output of multiple wind farms in the same region,a short-term power forecasting model of multiple wind farms based on the attention based spatial-tem-poral synchronous graph convolutional networks is proposed.First,an attention mechanism is introduced to quantify the impact of weather features on wind power,the local spatial-temporal graphs of wind power are constructed in three adja-cent time steps,and the local spatial-temporal features are extracted through convolutional calculation.Then,the input window’s overall spatial-temporal features are aggregated using the spatial-temporal synchronous graph convolutional layer.Finally,the power prediction results of multiple wind farms in the future are output through nonlinear mapping.The results of an actual example show that the proposed modelcan improve the day-ahead power forecasting accuracy of-multiple wind farms by 2.10%—13.94%through learning the dynamic spatial-temporal evolution law of wind power un-der different weather conditions.

关 键 词：深度学习 风电功率 相关性 时空同步图卷积网络 功率预测 

分 类 号：TM614[电气工程—电力系统及自动化]