基于随机森林与内核岭回归的配电网线损在线计算

Online Calculation of Distribution Network Line Loss Based on RF and KRR

针对配电网自动化水平低,其节点电气数据以及拓扑结构变化不能完全做到实时监控,使线损计算困难的问题,提出了基于随机森林(random forest,RF)以及内核岭回归(kernel ridge regression,KRR)的配电网线损在线计算方法。首先,根据配网历史运行情况确定不同拓扑结构下的典型运行模式;其次,通过对不同运行模式的配电系统进行潮流计算,分别构建训练集,通过对内核岭回归模型的训练,充分挖掘支路线损功率与配电网状态之间的映射关系;最后,基于随机森林分类法判断配电网运行模式,进而通过对应运行模式的内核岭回归模型实现配电网线损在线计算。对IEEE 33节点配电网的算例分析结果表明,所提方法具有较高的精度和较好的鲁棒性。

In view of the low automation level of distribution network,the changes of node electrical data and topology structure can⁃not be fully monitored in real time,which makes the line loss calculation difficult.This paper proposes an online calculation method of distribution network line loss based on random forest(RF)and kernel ridge regression(KRR).Firstly,the typical operation modes under different topologies are determined according to the historical operation conditions of the distribution network.Secondly,by the power flow calculation of distribution systems with different operation modes,the training sets are constructed respectively.Through the training of KRR model,the mapping relationship between branch line loss power and distribution network state is fully explored.Finally,the operation mode of distribution network is judged based on the RF classification method,then the online calculation of distribution network line loss is realized by the KRR model of the corresponding operation mode.The example analysis results of IEEE 33-bus distribution network show that this method has higher accuracy and better robustness.