基于孤立森林、模态分解和神经网络的空间负荷态势感知  

Spatial Load Situation Awareness Based on Isolation Forest, Mode Decomposition and Neural Networks

作　　者：肖白[1] 周文凯 姜卓[2] XIAO Bai;ZHOU Wenkai;JIANG Zhuo(School of Electrical Engineering,Northeast Electric Power University,Jilin 132012,China;School of Computer Science and Technology,Beihua University,Jilin 132021,China)

机构地区：[1]东北电力大学电气工程学院,吉林省吉林市132012 [2]北华大学计算机科学技术学院,吉林省吉林市132021

出　　处：《电力系统自动化》2022年第18期190-198,共9页Automation of Electric Power Systems

基　　金：国家自然科学基金资助项目(51177009);吉林省产业创新专项基金资助项目(2019C058-7)。

摘　　要：精准的空间电力负荷态势感知能够为城市电网的优化规划提供科学指导。为此,提出了一种基于孤立森林、变分模态分解、多层感知机和门控循环单元(iForest-VMD-MLP-GRU)的空间电力负荷态势感知方法。在态势觉察阶段,运用孤立森林算法对电力地理信息系统中既定空间分辨率下的Ⅰ类元胞负荷实测数据的异常值进行识别,并采用拉格朗日内插值法对其进行修正,从而确定出合理的Ⅰ类元胞负荷数据;在态势理解阶段,对态势觉察到的Ⅰ类元胞负荷数据运用变分模态分解方法进行分解,得到不同中心频率的分量,并根据其能量值确定趋势分量和低频分量;在态势预测阶段,采用多层感知机和门控循环单元分别对趋势分量和低频分量进行预测,并将两个分量的预测结果进行反演重构,得到目标年的Ⅰ类元胞负荷态势感知结果,之后采用网格化技术将其转化为基于Ⅱ类元胞的结果。实例分析结果证明了所述方法的正确性与有效性。Accurate spatial load situation awareness (SLSA) can provide a scientific guidance for the optimal planning of urban power grid.This paper proposes an SLSA method based on isolation forest,variational mode decomposition,multilayer perceptron and gated recurrent unit (i Forest-VMD-MLP-GRU).In the stage of situation perception,the isolation forest algorithm is used to identify the outliers of the measured data of classⅠcell load under the given spatial resolution in the power geographic information system,and the Lagrangian interpolation method is used to modify it,so as to determine the reasonable classⅠcell load data.In the stage of situation comprehension,the perceived classⅠcell load data is decomposed by the variational mode decomposition method to obtain the components of different central frequencies.The trend component and low-frequency component are determined according to their energy values.In the stage of situation forecasting,the multilayer perceptron and the gated recurrent unit are used to forecast the trend component and the low-frequency component respectively,and the forecasting results of the two components are inversed and reconstructed to obtain the situation awareness results of classⅠcell loads in the target year.The grid technology is then used to transform the results of classⅠcells into the results based on classⅡcells.The results of example analysis prove that the proposed method is correct and effective.

关 键 词：空间电力负荷 态势感知 孤立森林 变分模态分解 门控循环单元 多层感知机 

分 类 号：TM715[电气工程—电力系统及自动化] TP18[自动化与计算机技术—控制理论与控制工程]