| 引用本文: | 顾洁,孟璐,朱曈彤,刘书琪,金之俭.数据驱动的无精确建模含源配电网无功运行优化[J].电力自动化设备,2021,41(1): |
| GU Jie,MENG Lu,ZHU Tongtong,LIU Shuqi,JIN Zhijian.Data-driven optimization for reactive power operation in source distribution network without accurate modeling[J].Electric Power Automation Equipment,2021,41(1): |
|
| 摘要: |
| 分布式光伏的接入使得配电网无功电压运行控制需求及解决措施与传统配电网差异较大。针对配电网测量设备安装不全、网架参数难以准确获取，无法进行精确数学建模的问题，提出了无精确建模的含分布式光伏的配电网电压优化控制模型。以节点电压合格为优化目标，使用高速公路神经网络拟合网架节点注入功率与关键节点电压之间的映射关系；考虑分布式光伏的出力约束，进而采用定向寻优策略和反馈机制对优化模型进行求解；通过改变分布式电源逆变器出力来控制电网电压，实现全局系统电压控制。以不同规模的配电网实际数据为例，验证了所提优化运行控制模型的有效性。对比分析了采用普通神经网络和高速公路神经网络的电压拟合精度及收敛速度，证明高速公路神经网络应用于解决无精确建模的多节点含源配电网无功运行问题，可以实现拟合精度和拟合速度的双重优化。 |
| 关键词: 配电网 数据驱动 分布式光伏 无功运行优化 高速公路神经网络 |
| DOI：10.16081/j.epae.202011031 |
| 分类号:TM712 |
| 基金项目:国家重点基础研究计划项目(2016YFB0900100)；上海市科委科研计划资助项目(18DZ1100303) |
|
| Data-driven optimization for reactive power operation in source distribution network without accurate modeling |
|
GU Jie, MENG Lu, ZHU Tongtong, LIU Shuqi, JIN Zhijian
|
|
Research Center for Big Data Engineering and Technologies, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
|
| Abstract: |
| The access of distributed photovoltaic makes the requirement and solution of reactive power and voltage operation control for distribution network different from those of traditional distribution network. Aiming at the problems of incomplete installation of distribution network measurement equipment, difficulty in obtaining grid parameters accurately, and inability to carry out accurate mathematical modeling, a voltage optimization control model of distribution network with distributed photovoltaic is proposed without accurate modeling. Taking the qualified node voltage as the optimization goal, the highway neural network is used to fit the mapping between the injected power of grid nodes and the voltage of key nodes. Considering the output constraints of distributed photovoltaic, the directional optimization strategy and feedback mechanism are used to solve the optimization model. By changing the inverter output of distributed generation to control the grid voltage, the global system voltage control is realized. Taking the actual data of different scales of distribution networks as example, the effectiveness of the proposed optimal operation control model is veri-fied. The voltage fitting accuracy and convergence speed of the common neural network and the highway neural network are compared and analyzed, which proves that the highway neural network can be used to solve the reactive power operation problem of the multi-node source distribution network without accurate modeling, and the double optimization of fitting accuracy and fitting speed can be realized. |
| Key words: distribution network data-driven distributed photovoltaic reactive power operation optimization highway neural network |
