基于改进双向长短期记忆神经网络的新型电力系统源网暂态交互模型构建方法研究  

A Source-network Transient Interaction Modeling Method for New Power System Based on Improved Bi-LSTM Algorithm

作　　者：吕佳欣 于继来[1] LÜJiaxin;YU Jilai(School of Electrical Engineering and Automation,Harbin Institute of Technology,Harbin 150001,Heilongjiang Province,China)

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,黑龙江省哈尔滨市150001

出　　处：《电网技术》2024年第12期4896-4907,I0018,I0017,共14页Power System Technology

基　　金：国家自然科学基金项目(51877049)。

摘　　要：新型电力系统背景下,电力电子化、物理与数字融合和宏-微观电网复杂耦合的特征使得电网暂态在理论分析、仿真处理、控制算法等方面发生不同程度变化。该文以暂态态势为切入点,聚焦源-网协调的交互特性,提出新型电力系统源-网交互模型智能构建方法。首先以构建框架为媒介系统性阐释建模的核心思想与技术规划,分别从暂态交互响应、暂态等效模型的处理及适应性研究来说明智能建模方法的核心。其次,提出建立表底层映射融合分析法以进行样本获取,并根据建模需求及样本规模选取合适的智能算法。最后,以含风电场站的WSCC 9节点系统为例,构建暂态模型。经过图像及评估指标的对照,模型效果理想且具备适应性,验证了所提思想的合理性及选用算法的有效性。Under the background of new power system,the electricity electronization,physical and digital fusion and complex interconnection of the power grid make the transient states change in many aspects such as information,mechanism,simulation,analysis and control in different degree.Taking transient states as the entry point,this paper focuses on the equivalent construction of source-network station model.A new intelligent construction method of source-network interaction model in power system is proposed.Firstly,it systematically explains the technical thoughts and researching plan of modeling through the framework.It explains the core of the intelligent modeling method from transient interaction response,processing transient equivalent model and adaptability research respectively.Secondly,it is proposed to establish the mapping fusion analysis method which combines surface layer with ground layer for sample acquisition.The appropriate intelligent algorithm is selected according to the modeling requirements and sample size.Finally,the WSCC 9-node system with wind farm station is taken as the example to construct the model and verify the proposed method.

关 键 词：新型电力系统 简化模型 新能源场站 暂态稳定 深度学习 

分 类 号：TM721[电气工程—电力系统及自动化]