引用本文: | 叶睿恺,王慧芳,张森,张亦翔.数据驱动的含IIDG配电网短路电流计算多输出模型[J].电力自动化设备,2022,42(9): |
| YE Ruikai,WANG Huifang,ZHANG Sen,ZHANG Yixiang.Data-driven multi-output model for short circuit current calculation in distribution network with IIDGs[J].Electric Power Automation Equipment,2022,42(9): |
|
摘要: |
“双碳”目标下大量逆变型分布式电源(IIDG)接入配电网,使基于机理建模的配电网短路电流计算中,计算速度与准确性间的矛盾日益突出。数据驱动建模方法已被证实能有效解决两者间的矛盾,但采用单输出模型难以满足实际应用中要输出多个计算点的需求,容易产生模型数量问题。针对上述问题,提出了配电网短路电流多输出回归计算模型与计算方法。对问题转化与算法适应这2类多输出模型进行了分析对比,提出基于多目标回归模型融合方法、回归链方法等问题转化方法以及神经网络多输出方法,适合解决含IIDG配电网短路电流计算问题。分析了输入特征选择、性能评价指标、计算流程以及超参数寻优方法等关键问题。算例表明多输出模型能同时满足计算准确性和计算速度要求,性能强于单输出模型,且避免了模型数量问题。 |
关键词: 短路电流计算 多输出回归 逆变型分布式电源 配电网 数据驱动建模 |
DOI:10.16081/j.epae.202205015 |
分类号:TM713 |
基金项目:国家自然科学基金联合基金资助项目(U2166204) |
|
Data-driven multi-output model for short circuit current calculation in distribution network with IIDGs |
YE Ruikai, WANG Huifang, ZHANG Sen, ZHANG Yixiang
|
College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China
|
Abstract: |
A large number of IIDGs(Inverter Interfaced Distributed Generators) are connected to the distribution network under the “dual-carbon” target, which leads to an increasingly serious contradiction between the calculation speed and accuracy in short circuit current calculation of distribution network based on mechanism modeling method. Data-driven modeling method has been proved to be able to solve the contradiction effectively. However, single-output model is difficult to meet the demand of multiple calculation output in practical applications, which may cause a problem of model quantity. Aiming at the above problem, the multi-output regression modeling and calculation method for short circuit current in distribution network is proposed. The problem transformation method and algorithm adaptation method are analyzed and compared, the MTRS(Multi-Target Regressor Stacking),RC(Regressor Chains) and other problem transformation methods and multi-output neural network method are proposed, which are suitable for solving the short circuit current calculation problem of distribution network with IIDGs. The pivotal issues such as input feature selection, performance evaluation index, calculation process and optimization method of hyper parameters are analyzed. The example shows that the multi-output modeling method can satisfy calculation speed and accuracy simultaneously, and the performance of multi-output model is better than that of single-output model, and the problem of model quantity in single-output modeling method is avoided. |
Key words: short circuit current calculation multi-output regression inverter interfaced distributed generator distribution network date-driven modeling |