| 引用本文: | 张雨润,石重托,姚伟,黄伟,翟苏巍,文劲宇.融合电力系统知识的预想扰动下频率指标智能预测方法[J].电力自动化设备,2025,45(6):148-155,163. |
| ZHANG Yurun,SHI Zhongtuo,YAO Wei,HUANG Wei,ZHAI Suwei,WEN Jinyu.Intelligent prediction method of frequency indicators under anticipated disturbances integrated with power system knowledge[J].Electric Power Automation Equipment,2025,45(6):148-155,163. |
|
| 摘要: |
| 在稳态下基于预想扰动预测频率指标，进而制定相应的预防措施，是保证频率稳定的重要手段。基于此，提出一种融合电力系统知识的预想扰动下频率指标智能预测方法，分析频率指标与预想扰动大小之间的近似线性关系，并依据这一关系将频率指标划分为线性部分和非线性部分。建立虚拟变量回归模型预测线性部分，并使用嵌入电网拓扑的图卷积网络预测非线性部分。通过加和两部分得到预测结果。基于IEEE 39节点系统和WECC 179节点系统进行算例测试，结果验证了所提数据-知识融合驱动方法相比知识和数据驱动方法能够在不明显增加预测时间的同时大幅提升预测的准确性与可靠性。 |
| 关键词: 频率稳定 频率指标预测 预想扰动 数据-知识融合驱动 虚拟变量回归 图卷积网络 |
| DOI：10.16081/j.epae.202503006 |
| 分类号:TM712.3 |
| 基金项目:中国南方电网云南电网有限责任公司科技项目(0500002022030301XT00090)；国家自然科学基金资助项目(U22B20111) |
|
| Intelligent prediction method of frequency indicators under anticipated disturbances integrated with power system knowledge |
|
ZHANG Yurun1, SHI Zhongtuo1, YAO Wei1, HUANG Wei2, ZHAI Suwei3, WEN Jinyu1
|
|
1.State Key Laboratory of Advanced Electromagnetic Technology, School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, Wuhan 430074, China;2.Kunming Power Supply Bureau of Yunnan Power Grid Co.,Ltd.,Kunming 650012, China;3.Yunnan Power Dispatching and Control Center, Kunming 650011, China
|
| Abstract: |
| It is an important way to ensure frequency stability by predicting frequency indicators based on anticipated disturbances in steady state and then formulating corresponding prevention measures. Based on this, an intelligent prediction method of frequency indicators under anticipated disturbances integrated with power system knowledge is proposed. The approximate linear relationship between the frequency indicators and the anticipated disturbance magnitude is analyzed, and based on this relationship, the frequency indicators are divided into a linear part and a nonlinear part. A dummy variable regression model is built to predict the linear part, and a graph convolutional network embedded with the grid topology is used to predict the nonlinear part. The prediction results are obtained by summing the two parts. Case studies based on IEEE 39-bus system and WECC 179-bus system, the results validate that the proposed data-knowledge fusion-driven method is able to significantly improve the accuracy and reliability of the prediction without noticeably increasing the prediction time compared to the knowledge-driven and data-driven methods. |
| Key words: frequency stability frequency indicator prediction anticipated disturbances data-knowledge fusion-driven dummy variable regression graph convolutional network |
