摘要
配电台区广泛部署了传感器、监测装置等低压电网量测设备,这些设备在功率数据采集和转换过程中不可避免地受到非线性干扰因素的影响,而传统方法仅通过数据清洗来解决这一问题,忽略了数据本身的概率分布特性,导致状态感知结果失真。因此,提出一基于高斯混合模型(Gaussian Mixture Module,GMM)与深度神经网络(Deep Neural Network,DNN)的低压电网量测设备运行状态感知方法。通过对量测设备的低压电网的功率值进行求解,并提取每个量测节点的概率分布特性,简化了处理流程并降低了数据复杂性,实现数据的降维与关键信息保留。将处理后的概率分布功率数据输入到深度神经网络中进行训练。深度神经网络凭借其强大的自动学习和非线性映射能力,能够快速准确捕捉功率数据中和运行状态有关的隐藏关联,实现对低压电网量测设备的运行状态实时高效精准感知。实验结果表明:所提方法估计值接近实际,最大误差3.5 kW以内,实时性和准确性优于其他方法,能够精准感知低压电网量测设备状态,助力电网智能化管理。
Low-voltage power grid measurement equipment such as sensors and monitoring devices arc widely deployed in the distribution area.These devices arc inevitably affected by nonlinear interference factors during power data acquisition and conversion.Traditional methods only solve this problem through data cleaning,ignoring the probability distribution characteristics of the data itself,resulting in distorted state perception results.Therefore,a low-voltage power grid measurement equipment operation status perception method based on Gaussian Mixture Module(GMM)and Deep Neural Network(DNN)was proposed.By solving the power value of the low-voltage power grid of the measuring equipment and extracting the probability distribution characteristics of each measuring node,the processing flow was simplified and the data complexity was reduced,achieving data dimensionality reduction and key information retention.The processed probability distribution power data was input into a deep neural network for training.Deep neural networks,with their powerful automatic learning and nonlinear mapping capabilities,can quickly and accurately capture hidden correlations related to power data and operating status,achieving real-time,efficient,and accurate perception of the operating status of low-voltage power grid measurement equipment.The experimental results showed that the proposed method’s estimated values arc close to actual ones,with a maximum error of less than 3.5 kW.and its real-time and accuracy arc superior to those of other methods.It can accurately perceive the status of low-voltage power grid measurement equipment and assist in intelligent management of the power grid.
作者
马聪智
魏然
杨涛
陈沼宇
白苏娜
MA Congzhi;WEI Ran;YANG Tao;CHEN Zhaoyu;BAI Su'na(State Grid Tianjin Electric Power Economic&Technology Research Institute,Tianjin 300171 China;State Grid Tianjin Electric Power Company,Tianjin 300000 China)
出处
《国外电子测量技术》
2025年第3期195-202,共8页
Foreign Electronic Measurement Technology
基金
国网天津市电力公司科技项目(经研-研发2024-01)。
关键词
配电台区
低压电网
运行状态感知
深度神经网络
概率分布特性
distribution station area
low-voltage power grid
perception of operational status
deep neural network
probability distribution characteristics
