基于数据驱动的新型电力系统潮流分析方法

New Power System Power Flow Analysis Method Based on Data-driven

为满足未来新型电力系统高精度实时潮流分析的需求,提出一种基于数据驱动的新型电力系统潮流分析方法。该方法可充分利用现有的海量数据,并应用数据驱动技术进行电力大数据挖掘。首先,基于传统潮流分析模型与源-网-荷运行特点,构建表征潮流分析的输入和输出特征;其次,根据系统规模及数据复杂度,在传统全连接深度神经网络基础上,选择部分隐含层进行随机失活设计,以提高深度学习的泛化性能;最后,基于某供电局110 kV电网历史运行的15 min级颗粒度数据,对设计的网络进行模型训练,并验证了数据驱动算法的有效性和可行性。

In order to meet the demand for high-precision and real-time power flow analysis of new power systems in the future,a new power system power flow analysis method is proposed based on data-driven.This method can make full use of the existing massive data,and applies data driven technology to power big data mining.Firstly,based on the traditional power flow analysis model and source-network-load operation characteristics,the input and output characteristics representing power flow analysis are constructed;Secondly,according to the system scale and data complexity,based on the traditional full connection deep neural network,some hidden layers are selected for random deactivation design to improve the generalization performance of deep learning;Finally,based on the 15 minute level granularity data of historical operation from a power supply bureau’s 110 kV grid,the designed network is trained,and the effectiveness&feasibility of data-driven algorithm is v erified.