基于数据驱动的高比例新能源配电网无功优化算法  

Data-driven reactive power optimization algorithm for the distribution network with high proportion of renewable energy

作　　者：林泓宏 余涛[1] 张桂源 张孝顺 LIN Honghong;YU Tao;ZHANG Guiyuan;ZHANG Xiaoshun(College of Electric Power,South China University of Technology,Guangzhou 510640,China;China Southern Power Grid Industry Investment Group Company Limited,Guangzhou 510630,China;Foshan Graduate School of Innovation,Northeastern University,Foshan 528311,China)

机构地区：[1]华南理工大学电力学院,广州510640 [2]南方电网产业投资集团有限责任公司,广州510630 [3]东北大学佛山研究生创新学院,广东佛山528311

出　　处：《综合智慧能源》2023年第11期10-19,共10页Integrated Intelligent Energy

基　　金：中央高校基本科研业务资助(N2229001)。

摘　　要：随着大量分布式新能源接入配电网,其电能损耗和节点电压偏差增大等问题愈发严重,仅依赖安装无功补偿设备、改变传统发电机机端电压等手段已无法满足高比例新能源配电网的高调节性能需求。为此,研究深度挖掘不同种类新能源对于配电网的无功调控潜力,详细分析了目标函数以及约束条件的设置,规划了基于数据驱动的无功调控算法流程,实现了合理调节配电网的节点电压与降低配电线路电能损耗的目标。首先,构建了高比例新能源参与配电网无功优化的数学模型,并应用多种智能算法求解。随后,采用多种算法得到的无功调控策略作为深度学习长短期记忆(LSTM)神经网络的训练数据,通过训练的网络在此类模型下可以预测得到优质的、高效率的无功调控策略。最后,利用IEEE 14与IEEE 33节点算例验证了所提算法的有效性和优化性能。With the integration of massive distributed new energy into distribution networks,power loss and voltage deviation at nodes are becoming increasingly severe.The high requirements on high-proportion renewable energy distribution networks'adjustability can not be satisfied by traditional measures merely,such as installation of reactive power compensation devices,adjusting the voltage at the generator side.Therefore,the reactive power regulation capacities of different types of new energy generators in distribution networks are study deeply.The study analyses the objective function and constraint settings,and outlines a data-driven reactive power regulation algorithm.The algorithm can regulate node voltage and reduce power loss of the distribution network at the same time.Firstly,a mathematical model of reactive power optimization for the high-proportion new energy distribution network is constructed,and solved by various intelligent algorithms.Subsequently,the reactive power regulation solution sets obtained from the algorithms are used as training data for the deep learning long-short-term memory(LSTM)network.The trained network can predict reactive power regulation strategies efficiently based on the model above.Finally,the effectiveness and optimization performance of the proposed algorithm is validated by in an IEEE 14-bus system and an IEEE 33-bus system.

关 键 词：高比例新能源 配电网 深度学习 无功优化 帕累托前沿 

分 类 号：TK01[动力工程及工程热物理]