基于支持向量回归数据驱动的配电网潮流回归

Date-driven Power Flow Regression Using SVR in Distribution Grid

随着大量分布式电源接入配电网,配电网潮流分布由原来的单向流动转为双向流动,配电网潮流分布不均、电压越限等问题频现,研究适用于主动配电网的潮流计算更为重要.由于中低压配电网电气参数往往收集不到,量测系统也不如输电网完备,开关动作导致的拓扑变化难以实时反映到监控系统.实践中难以将基于导纳矩阵的输电网潮流方法应用到中低压配电网中.鉴于此,本文提出一种数据驱动的潮流线性回归模型,以实现不依赖于配电网物理模型的潮流计算与分析.首先建立不同类型的母线已知量与未知量的映射关系;其次,进一步推导该模型母线类型变换的更新方式;然后构建基于支持向量回归(support vector regression,SVR)的潮流回归模型,通过嵌入高斯核函数以及对样本进行聚类更好地拟合潮流的非线性;最后,在多个IEEE标准系统和改进的IEEE33节点系统仿真验证了所提方法的有效性.

With the connection of a large number of distributed generators to distribution grid,the power flow distribution has changed from the original one-way flow to the two-way flow,and the problems of uneven power flow distribution and the over-limit voltage in distribution grid have frequently appeared.Therefore,it is more important to study the power flow calculation applicable to the active distribution grid.Because the electrical parameters of the medium and low voltage distribution grid are often not collected,and the measurement system is not as complete as transmission network,the topological changes caused by the switching action are difficult to reflect to the monitoring system in real time.In practice,it is difficult to apply the power flow method of transmission grid based on admittance matrix to medium and low voltage distribution grids.In view of this point,a data-driven linear regression model of power flow to realize the calculation and analysis of power flow independent of the physical model of the distribution grid is proposed in this paper.Firstly,the mapping relationship between the known quantity and the unknown quantity of different types of bus bars is established.Secondly,the update method of the bus type transformation of the model is further derived.Then,a power flow regression model based on support vector regression(SVR)is constructed.The non-linear power flow is better fitted by embedding Gaussian kernel function and clustering samples.Finally,simulations on multiple IEEE standard systems and improved IEEE33 node systems verify the effectiveness of the proposed method.