| 引用本文: | 赵冬梅,谢家康,杜泽航,魏中庆,田世芳,徐咏盛.基于统计信息系数和Wasserstein生成对抗网络的风火系统暂态特征选择与两阶段稳定评估[J].电力自动化设备,2023,43(4): |
| ZHAO Dongmei,XIE Jiakang,DU Zehang,WEI Zhongqing,TIAN Shifang,XU Yongsheng.Transient feature selection and two-stage stability assessment of wind-fire system based on uniform information coefficient and Wasserstein-generative adversarial network[J].Electric Power Automation Equipment,2023,43(4): |
|
| 摘要: |
| 为实现含新能源电力系统暂态稳定的在线更新，降低新能源机组出力、电压等特征对评估准确率的影响，提出基于统计信息系数和Wasserstein生成对抗网络的暂态特征选择和两阶段稳定评估模型。首先，改进最大信息系数中的网格划分方法，建立基于Pearson相关系数和统一信息系数的混合Markov决策模型，对风火系统暂态特征进行数据分析与特征选择。其次，以Wasserstein距离代替传统生成对抗网络中的JS散度，提出基于Wasserstein生成对抗网络的暂态稳定模型在线更新方法。同时依据一阶段暂态稳定评估置信度将样本划分为安全域样本与非安全域样本，提出一种两阶段暂态稳定评估方法。最后在改进的IEEE 39节点、IEEE 118节点系统中将所提方法与传统方法进行了对比，结果表明所提方法的暂态稳定评估准确度有所提高，且计算速度有所加快。 |
| 关键词: 暂态特征选择 Wasserstein生成对抗网络 统计信息系数 Markov决策模型 两阶段暂态稳定评估 |
| DOI：10.16081/j.epae.202211001 |
| 分类号:TM712 |
| 基金项目:国家重点研发计划项目(2017YFB0902600)；国家电网公司科技项目(SGJS0000DKJS1700840) |
|
| Transient feature selection and two-stage stability assessment of wind-fire system based on uniform information coefficient and Wasserstein-generative adversarial network |
|
ZHAO Dongmei1, XIE Jiakang1, DU Zehang1, WEI Zhongqing1, TIAN Shifang2, XU Yongsheng3
|
|
1.School of Electrical and Electronic Engineering, North China Electric Power University, Beijing 102206, China;2.Tongshan Power Supply Company of State Grid Hubei Electric Power Company, Xianning 437600, China;3.Changchun Power Supply Company of State Grid Jilin Electric Power Company, Changchun 130021, China
|
| Abstract: |
| In order to realize the online updating of the transient stability of the power system containing new energy and reduce the influence of new energy units’ features on the assessment accuracy such as the power output and the voltage, a transient feature selection and two-stage stability assessment model based on uniform information coefficient(UIC) and Wasserstein-generative adversarial network(W-GAN) is proposed. Firstly, the mesh partition method in the maximum information coefficient is improved, and the hybrid Markov decision model based on Pearson relative coefficient and UIC is established to analyze the data and select the transient feature of the wind-fire system. Secondly, the Jensen-Shannon(JS) divergence in the traditional generative adversarial network is replaced by Wasserstein distance. An on-line updating method of transient stability model based on W-GAN is proposed. At the same time, according to the confidence of one-stage transient stability assessment, the samples are divided into safe domain samples and non-safe domain samples, and a two-stage transient stability assessment method is proposed. Finally, the proposed method is compared with the traditional method in the improved IEEE 39-bus system and IEEE 118-bus system. The results show that the accuracy of transient stability assessment of the proposed method is improved and its calculation speed is accelerated. |
| Key words: transient feature selection Wasserstein-generative adversarial network uniform information coefficient Markov decision model two-stage transient stability assessment |
