基于WKNN和KELM-GPR的虚拟电厂交互模型构建方法  

CONSTRUCTION METHOD OF VIRTUAL POWER PLANT INTERACTION MODEL BASED ON WKNN AND KELM-GPR

作　　者：郝瑞鑫 樊艳芳[1] 侯俊杰[1] 白雪岩 刘雨佳 Hao Ruixin;Fan Yanfang;Hou Junjie;Bai Xueyan;Liu Yujia(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China)

机构地区：[1]新疆大学电气工程学院,乌鲁木齐830047

出　　处：《太阳能学报》2024年第12期637-649,共13页Acta Energiae Solaris Sinica

基　　金：新疆维吾尔自治区自然科学基金(2022D01C365,2022D01C662);2022天山英才培养计划(2022TSYCLJ0019)。

摘　　要：目前虚拟电厂参与配电网调度多依赖物理模型,然而,由于虚拟电厂聚合成员的多元性、时变性、时序耦合性,导致其解析建模难度增大,难以满足配电网日内调度需求的时效性,且存在隐私安全问题。因此,提出一种基于加权K最近邻(WKNN)和核极限学习机-高斯过程回归(KELM-GPR)的虚拟电厂交互模型构建方法。首先为提升交互模型的预测精度,提出一种均匀生成训练集的方法;其次通过WKNN算法建立调度指令可行性模型,衡量虚拟电厂的可调度边界;接着引入GPR作为误差补偿模型,并与KELM结合,构建基于KELM-GPR的虚拟电厂交互成本模型,以参与配电网的经济调度;最后为验证所提方法的可行性,基于虚拟电厂调度指令可行性和交互成本模型,构建虚拟电厂参与配电网日内优化调度模型。仿真结果表明,所提方法能显著减少模型优化求解时间,并能保护虚拟电厂内部信息安全。Currently,the participation of the virtual power plant in distribution network dispatching mostly depends on physical model.However,due to the diversity,time-varying,and temporal coupling of the aggregated members of virtual power plant,its analytical modeling becomes more difficult,and it is difficult to meet the timeliness of the intra-day scheduling requirements of distribution network,and there are privacy security issues.Therefore,a virtual power plant interaction model construction method based on weighted K-nearest neighbor(WKNN)and kernel extreme learning machine-Gaussian process regression(KELM-GPR)is proposed.Firstly,to improve the prediction accuracy of the interactive model,a method of uniformly generating training sets is proposed;Secondly,a scheduling instruction feasibility model is established through the WKNN algorithm to measure the dispatchable boundary of the virtual power plant;Next,GPR is introduced as an error compensation model,and combined with KELM to construct a virtual power plant interaction cost model based on KELM-GRP to participate in the economic dispatch of the distribution network;Finally,to verify the feasibility of the proposed method,based on the model of scheduling instruction feasibility and interactive cost of virtual power plant,a virtual power plant participation in the intra-day optimization scheduling of the distribution network model is constructed.Simulation results show that the proposed method can significantly reduce the model optimization solution time and protect the internal information security of the virtual power plant.

关 键 词：虚拟电厂 机器学习 优化调度 误差补偿 建模方法 

分 类 号：TM73[电气工程—电力系统及自动化]