摘要
大规模风电接入给电网的规划和运行带来了极大的不确定性,导致系统运行模式更加多变。极端运行场景提取对于分析电网运行的高风险薄弱点具有重要意义,而传统经验场景提取方式难以应对风电和负荷的双重不确定性。面向规划层面的电力系统安全稳定评估问题,提出一种基于数据挖掘和机器学习算法的电网极端运行场景提取方法。首先,通过机器学习识别出对暂态功角稳定影响较大的场景变量并依据重要程度进行排序,同时采用熵权法体现场景变量自身的离散性对极端运行场景的贡献程度。随后,利用加权聚类算法筛选出代表大多数场景暂态功角稳定水平的典型运行场景,进而提取出离群、极端的边缘运行点作为极端场景。最后,采用IEEE 39节点算例进行暂态仿真分析,验证了采用数据挖掘与具体问题相结合的方法进行极端场景提取的有效性和合理性,提升了风电并网规划、稳定分析的水平和效率。
The large-scale integration of wind power brings great uncertainty to the operation and planning of power grids, making the system operation mode more variable. The extraction of extreme operation scenarios is of great significance for analyzing the high-risk and weak points in the power grid operation, while the traditional empirical scenario extraction methods have difficulty in dealing with the dual uncertainties of wind power and load. Aiming at the security and stability assessment problem of the power system at the planning level, a method for extracting extreme operation scenarios in the power grid based on data mining and machine learning algorithms is proposed. First, through machine learning, the scenario variables that have a greater impact on the transient stability are identified and sorted according to their importance. At the same time, the entropy-weighted method is used to reflect the degree of contribution of the discreteness of the scenario variable itself to the extreme operation scenarios. Then, the weighted clustering algorithm is used to pick the typical operation scenarios that can represent the transient angle stability level in most scenarios. Furthermore, the outlier and extreme edge operating points far from the clustering center are extracted as extreme scenarios. Finally, an IEEE 39-bus test system is used for the transient simulation analysis, which verifies the effectiveness and rationality of the method that combining data mining and specific problems in extracting extreme scenarios. The proposed method improves the level and efficiency of grid integration planning and stability analysis of wind power.
作者
罗魁
石文辉
LUO Kui;SHI Wenhui(China Electric Power Research Institute,Beijing 100192,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2021年第20期113-120,共8页
Automation of Electric Power Systems
基金
中国电力科学研究院创新基金项目(NY83-19-003)资助。
关键词
赋权聚类
机器学习
熵权法
极端场景
暂态功角稳定
weighted clustering
machine learning
entropy-weighted method
extreme scenario
transient angle stability
