| 引用本文: | 邢 超,魏荣智,奚鑫泽,等.基于GRU的特高压三端混合直流输电线路故障区域识别方法[J].电力系统保护与控制,2023,51(10):98-109.[点击复制] |
| XING Chao,WEI Rongzhi,XI Xinze,et al.Fault zone identification method for three-terminal hybrid UHVDC transmission lines based on GRU[J].Power System Protection and Control,2023,51(10):98-109[点击复制] |
|
| |
|
|
| 本文已被：浏览 5911次 下载 1232次 |
码上扫一扫！ |
| 基于GRU的特高压三端混合直流输电线路故障区域识别方法 |
| 邢超,魏荣智,奚鑫泽,刘明群,何鑫,李胜男 |
|
|
| (云南电网有限责任公司电力科学研究院，云南 昆明 650217) |
|
| 摘要: |
| 针对特高压三端混合直流输电线路故障区域识别问题，提出一种基于门控循环单元(gate recurrent unit, GRU)的特高压三端混合直流输电线路故障区域识别方法。首先，分析了直流线路昆北侧边界和龙门侧边界直流功率传递函数的幅频特性、T区线模功率突变量的正负差异，指出三端直流线路不同区域故障时的故障特征差异。其次，对线模功率进行多尺度小波分解，提取线模功率高频能量，结合T区线模功率突变量、正负极功率突变量幅值，组成故障特征量，作为GRU的输入量，故障区域作为输出量，构建GRU故障区域识别模型。然后，将测量点得到的故障特征量输入训练完成的GRU模型中，即可达到故障区域识别的目的。通过大量仿真，验证了所提故障区域识别方法准确率高，且可耐受300 Ω的过渡电阻。 |
| 关键词: 特高压三端混合直流 幅频特性 门控循环单元 故障特征量 故障区域识别 |
| DOI：10.19783/j.cnki.pspc.221174 |
| 投稿时间：2022-07-24修订日期：2022-10-31 |
| 基金项目:国家自然科学基金项目资助(51707026) |
|
| Fault zone identification method for three-terminal hybrid UHVDC transmission lines based on GRU |
| XING Chao,WEI Rongzhi,XI Xinze,LIU Mingqun,HE Xin,LI Shengnan |
| (Electric Power Research Institute of Yunnan Power Grid Co., Ltd., Kunming 650217, China) |
| Abstract: |
| There is a problem of fault area identification of UHV tri-terminal hybrid HVDC transmission lines. Thus this paper proposes a fault zone identification method for three-terminal hybrid UHVDC transmission lines based on GRU. First, this paper analyzes amplitude-frequency characteristics of the DC power transfer function of the Kunbei and Longmen boundaries, and positive and negative difference of line mode power mutation in the T-zone. Also the difference of the fault characteristics when the fault occurs at different zones of three-terminal DC lines is found. Second, the wavelet transform is used to perform multi-scale wavelet decomposition of line mode power, the high frequency energy is extracted. Combined with line mode power mutation in the T-zone and the positive and negative DC power changes, the fault characteristics are formed and used as the input quantity of GRU. The fault area is used as the output quantity to construct the GRU fault area identification model. Then, the fault zone can be obtained by inputting fault characteristics to the trained GRU model. A large number of simulations verify that the method proposed has high accuracy and can withstand a transition resistance of 300 Ω. |
| Key words: three-terminal hybrid UHVDC amplitude-frequency characteristic GRU fault characteristic data fault zone identification |
|
|
