| 引用本文: | 蔡昌春,息梦蕊,刘昊林,陈洁,赵东.基于数据驱动和多场景技术的微电网并网等效建模[J].电力自动化设备,2022,42(9): |
| CAI Changchun,XI Mengrui,LIU Haolin,CHEN Jie,ZHAO Dong.Grid-connected equivalent modeling of microgrids based on data-driven and multi-scenario technologies[J].Electric Power Automation Equipment,2022,42(9): |
|
| 本文已被：浏览 3340次 下载 1190次 |
|
|
|
| 基于数据驱动和多场景技术的微电网并网等效建模 |
|
蔡昌春1,2, 息梦蕊1,2, 刘昊林1,2, 陈洁1,2, 赵东3
|
|
1.河海大学 江苏省输配电装备技术重点实验室，江苏 常州 213022;2.河海大学 物联网工程学院，江苏 常州 213022;3.国网马鞍山供电公司，安徽 马鞍山 243000
|
|
| 摘要: |
| 微电网并网等效建模是含大规模微电网并网电力系统仿真分析的基础，但分布式电源的随机性和波动性导致微电网并网模型参数的不确定性和结构的多样性。提出一种基于数据驱动和多场景技术的微电网并网等效建模方法，利用马尔可夫链建立基于分布式电源随机概率特性的微电网运行场景，明晰微电网场景间的时间关联性。在传统k-means聚类算法的基础上增加马尔可夫影响因子，加速微电网运行场景向聚类中心聚集以及场景消减，从而获得微电网典型运行场景。针对每个典型运行场景建立基于长短期记忆(LSTM)神经网络的微电网并网等效模型，利用LSTM神经网络的时序逻辑特性和内部非线性映射特性描述微电网并网点整体动态运行特性。仿真结果验证了所提方法的合理性和有效性。 |
| 关键词: 微电网 等效建模 多场景技术 k-means聚类算法 LSTM神经网络 |
| DOI：10.16081/j.epae.202203009 |
| 分类号:TM732 |
| 基金项目:国家自然科学基金资助项目(51607057)；中央高校基本科研业务费专项资金资助项目(2020B22514)；江苏省输配电装备技术重点实验室开放基金资助项目(2021JSSPD07) |
|
| Grid-connected equivalent modeling of microgrids based on data-driven and multi-scenario technologies |
|
CAI Changchun1,2, XI Mengrui1,2, LIU Haolin1,2, CHEN Jie1,2, ZHAO Dong3
|
|
1.Jiangsu Key Laboratory of Power Transmission & Distribution Equipment Technology, Hohai University, Changzhou 213022, China;2.College of the IOT Engineering, Hohai University, Changzhou 213022, China;3.State Grid Ma’anshan Power Supply Campany, Ma’anshan 243000, China
|
| Abstract: |
| The grid-connected equivalent modeling of microgrids is the basis of simulation and analysis for power system connected with large scale microgrids, while the randomness and fluctuation of distributed gene-ration cause uncertain parameters and diverse structures of the grid-connected model of microgrids. A grid-connected equivalent modeling method of microgrids is proposed based on data-driven and multi-scenario technologies, Markov chain is used to build the operation scenarios of microgrids based on the random probability characteristics of distributed generation, and the temporal correlation among the scenarios of micro-grids is defined. Markov influence factor is added based on the traditional k-means clustering algorithm to accelerate the gathering of operation scenarios of microgrids to clustering centers and the scenario cutting, thus the typical operation scenarios of microgrids are obtained. The grid-connected equivalent model of microgrids based on LSTM(Long Short-Term Memory) neural network is built for each typical operation scenario, and the temporal logic characteristics and internal non-linear mapping characteristics of LSTM neural network are used to describe the whole dynamic operation characteristics of microgrids at the points of common coupling. Simulative results verify the rationality and validity of the proposed method. |
| Key words: microgrid equivalent modeling multi-scenario technology k-means clustering algorithm LSTM neural network |
