基于特征线性调制深度学习网络的馈线负荷构成特征识别

Feature recognition of feeder load composition based on deep learning network based on feature linear modulation

随着电网规模越来越大,结构越来越复杂,供电可靠性的要求越来越高,通过现场试验的方式获取电力负荷的动态特性变得十分困难。提出了基于统计综合法的思想,运用电力系统仿真软件,通过仿真计算的方式得出对应电力负荷的动态特性。为了获得馈线负荷构成比例识别所需的负荷数据库,建立了包含变频空调负荷、分布式光伏负荷、感应电机负荷以及静态负荷的电网模型,通过设置电压跌落,得到不同负荷情况馈线功率数据,建立馈线负荷数据库,运用基于特征线性调制的改进深度学习网络的方法挖掘负荷节点或馈线功率变化的特征,并对负荷构成进行识别。结果经过训练后,与传统BP神经网络进行对比,识别准确率达到99%以上,部分负荷可以达到99.6%以上,有着较高的识别准确度。

With the increasing scale of power grid,the structure is more and more complex,and the requirement of power supply reliability is higher and higher,it is very difficult to obtain the dynamic characteristics of power load through field test.Based on the idea of statistical synthesis method,this paper uses the power system simulation software to obtain the dynamic characteristics of the corresponding power load through simulation calculation.In order to obtain the load database required for the identification of feeder load composition proportion,the power grid model including variable frequency air conditioning load,distributed photovoltaic load,induction motor load and static load is established as shown in the figure.By setting voltage drop,the feeder power data under different load conditions are obtained,and the feeder load database is established,The improved deep learning network method based on feature linear modulation is used to mine the characteristics of load node or feeder power change,and identify the load composition.Following training,the results were compared to those of the traditional BP neural network,revealing a recognition accuracy surpassing 99%.Additionally,certain loads exhibited accuracy exceeding 99.6%,thus indicating a significantly elevated level of recognition precision.