基于时空库普曼自动编码器的风电场短期风速预测
Short-term Wind Speed Prediction in Wind Farms Based on Spatiotemporal Koopman Autoencoder
王轶琳 1刘丰瑞 2李宗锴1
作者信息
- 1. 东北电力大学吉林 吉林 132012
- 2. 澳门大学,澳门特别行政区 999078
- 折叠
摘要
为了提升以新能源为主体的新型电力系统的风电消纳水平,需要对风速进行精确预测,关键在于提炼风电系统动态趋势与风速序列中潜在的物理结构.依据库普曼动力学理论与自编码器思想搭建物理约束时空神经网络,生成风电场非线性变量的线性演化矩阵.首先,通过线性演化矩阵近似系统趋势,在预测过程中充分考虑前、后向动态.然后,设置双向相关预测机制与适配不同对象的代价函数,降低预测对序列的可逆性、平稳性要求.同时,对特征空间隐向量进行可视化,展现系统内特征区间依赖.最后,借助北票王子山风电场风速实测数据验证所提方法的有效性.结果表明:所提方法对于强随机、强波动的风速序列具有较高的预测精度,泛化能力强,可解释性优越.
Abstract
In order to improve the wind power consumption level of the new power system dominated by new energy,it is necessary to accurately predict wind speed,and the key is to extract the dynamic trend of the wind power system and the potential physical structure in the wind speed sequence.Build a physically constrained spatiotemporal neural network based on the Kupfmann dynamics theory and autoencoder idea,and generate a linear evolution matrix for the nonlinear variables of the wind farm.Firstly,approximate the system trend through a linear evolution matrix and fully consider the forward and backward dynamics in the prediction process.Then,a bidirectional correlation prediction mechanism and cost functions adapted to different objects are set up to reduce the reversibility and stationarity requirements of the prediction on the sequence.At the same time,visualize the hidden vectors in the feature space to show the dependency of feature intervals within the system.Finally,the effectiveness of the proposed method was verified using the measured wind speed data from the Beipiao Wangzi Mountain Wind Farm.The results show that the proposed method has high prediction accuracy,strong generalization ability,and superior interpretability for wind speed sequences with strong randomness and strong fluctuations.
关键词
短期风速预测/新型电力系统/库普曼理论/时空神经网络/物理约束学习Key words
short-term wind speed prediction/new power system/Koopman theory/spatiotemporal neural networks/physical constraint learning引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据