面向高比例新能源接入的配电网电压时空分布感知方法

Perception Method of Voltage Spatial-Temporal Distribution of Distribution Network with High Penetration of Renewable Energy

随着配电网中新能源渗透率不断增加,电压越限、潮流过载等安全性问题愈发严峻。为此,文中提出一种面向高比例新能源渗透的配电网电压时空分布感知方法,通过数据驱动的方法实现在缺乏配电网潮流模型条件下的短期、高精度的节点电压感知预测。所提方法包含三部分:基于数值气象预报的分布式风光预测,建立气象数据与分布式能源出力之间的时空映射关系;基于广义回归神经网络(GRNN)的电压灵敏度矩阵学习机制,在缺乏配电网潮流模型条件下构建数据驱动的节点功率-电压映射;基于核密度估计(KDE)的GRNN预测样本修正法(KDE-GRNN),进一步降低因原始样本局部密度偏差导致的预测误差。基于IEEE 33和委内瑞拉141节点配电网验证了所提方法的有效性。对比同类算法,验证了提出的KDE-GRNN在预测精度、收敛速度方面的优势。

With the increasing penetration of renewable energy in distribution networks,a series of reliability issues,e.g.,voltage over-limit and power flow overloading,have become severe threats.Therefore,an perception method of voltage spatial-temporal distribution with high penetration of renewable energy is proposed.Due to the absence of power flow models of the distribution network in practice,a data-driven method is designed to make accurate short-term prediction for nodal voltage perception.The proposed method is composed of three sectors:numerical weather prediction(NWP)based distributed wind power and photovoltaic forecasting,in which the relationship between meteorological data and distributed energy output is developed;generalized regression neural network(GRNN)based learning mechanism for constructing voltage sensitivity matrix,in which data-driven nodal power-voltage mapping is developed without a power flow model of the distribution network;and the kernel density estimation(KDE)based GRNN sample amendment method for avoiding the forecasting errors caused by the local density deviation of the original sample.Case studies based on IEEE 33-bus and Venezuela 141-bus distribution systems demonstrate the effectiveness of the proposed method.Compared with similar methods,the proposed KDE-GRNN has a significant advantage in forecast precision and rate of convergence.