| 引用本文: | 杨大坤,张大海,严嘉豪,等.计及源-荷-环境之间多影响因素耦合的风-光-荷联合场景生成[J].电力系统保护与控制,2025,53(23):63-74.[点击复制] |
| YANG Dakun,ZHANG Dahai,YAN Jiahao,et al.Multi-factor coupled scenario generation for wind-solar-load systems considering source-load-environment interactions[J].Power System Protection and Control,2025,53(23):63-74[点击复制] |
|
| 摘要: |
| 在电力系统能源快速转型的背景下，高比例可再生能源和多元负荷的接入加剧了电网运行场景的复杂性，源-荷-环境间多影响因素的深度耦合增加了运行场景特征提取难度。针对上述问题，提出一种综合考虑复杂环境耦合和多维时序特征的风-光-荷场景生成方法。首先，考虑到风-光-荷的强时序耦合与非线性特性，采用双向门控循环单元(bidirectional gated recurrent unit, Bi-GRU)对环境因素和风-光-荷数据的初始特征进行提取，并通过误差迭代技术优化特征向量。其次，为精确捕捉风-光-荷及其环境因素之间的概率依赖关系，将提取的初始特征作为输入，并基于条件变分自编码器(conditional variational auto encoder, CVAE)构建风-光-荷场景生成模型。然后，针对风-光-荷功率曲线的时间序列非等长问题，采用动态时间规整(dynamic time warping，DTW)算法对生成的风-光-荷场景进行约简，确保最终生成的风-光-荷场景具有时序一致性。最后，基于中国某地区实测的风-光-荷数据，对所提方法进行验证。结果表明，该方法能够生成更加接近实际情况的风-光-荷联合场景。 |
| 关键词: 时序耦合 特征提取 条件变分自编码器 风-光-荷场景生成 |
| DOI：10.19783/j.cnki.pspc.250197 |
| 投稿时间：2025-02-27修订日期：2025-08-27 |
| 基金项目:国家重点研发计划项目资助(2022YFB2403400) |
|
| Multi-factor coupled scenario generation for wind-solar-load systems considering source-load-environment interactions |
| YANG Dakun1,ZHANG Dahai1,YAN Jiahao2,LI Yaping2,SUN Kai1,JIAO Zhijie1 |
| (1. School of Electrical Engineering, Beijing Jiaotong University, Beijing 100044, China;
2. China Electric Power Research Institute, Nanjing 210003, China) |
| Abstract: |
| With the rapid energy transition in power systems, the growing penetration of renewable energy and diversified loads has increased the complexity of grid operating scenarios. The deep coupling of multiple influencing factors among energy sources, loads, and environmental conditions further complicates scenario feature extraction. To address these challenges, this paper proposes an integrated wind-solar-load scenario generation method incorporating complex environmental coupling and multi-dimensional temporal features. First, considering strong temporal coupling and nonlinear characteristics of wind, solar, and load profiles, a bidirectional gated recurrent unit (Bi-GRU) is used to extract initial features from environmental factors and historical data, with feature vectors refined via error iteration. Subsequently, to accurately capture the probabilistic dependencies among wind, solar, load, and environmental variables, the extracted initial features are fed into a conditional variational autoencoder (CVAE) to construct a wind-solar-load scenario generation model. Then, to address the issue of unequal sequence lengths in wind-solar-load power curves, the dynamic time warping (DTW) algorithm is employed to reduce the generated scenarios, ensuring temporal consistency across the final output. Finally, the proposed method is validated using real-world wind-solar-load data from a region in China. The results demonstrate that the method can generate joint scenarios that more closely reflect real-world operating conditions. |
| Key words: temporal coupling feature extraction conditional variational autoencoder wind-solar-load scenario generation |
