标幺化特征值的负荷阈值辨识方法研究

Research on load threshold identification method for per-unit eigenvalues

为降低非侵入式负荷在线辨识算法对硬件资源的要求,文中提出一种基于电流稳态特性分析的负荷特征标幺化阈值辨识方法。该方法将单个电器设备与多个电器设备的每种工作模态均视为不同负荷类别,应用瞬时功率理论与傅里叶变换方法提取不同模态基波有功、无功与各频次谐波电流分量幅值,并将其作为负荷特征;再以不同负荷类别的标准特征值为基准,提高负荷辨识中谐波电流特征权重,应用所提阈值辨识方法对待辨识负荷类别作出判定。结果表明:离线状态下,文中方法与k-NN、BP神经网络两种方法的辨识准确率相差1%左右;在线状态下,该方法的辨识准确率接近90%,且采用DSP28335处理器完成一次辨识任务耗费的时间在10 ms以内。辨识准确率与辨识时效性两项指标良好,证实了所提方法有效可行。

A load characteristics per-unit threshold identification method based on the analysis of current steady-state characteristics is proposed to reduce the hardware resource requirements of the non-intrusive online load identification algorithm.In this method,each operating mode of single electrical equipment and multiple electrical equipments are regarded as different load categories,and instantaneous power theory and Fourier transform method are used to extract the amplitude of fundamental wave active power,reactive power of different modes and harmonic current components of different frequencies and make it as load characteristics.The standard characteristic value of different load categories is taken as a benchmark to increase the weight of harmonic current characteristics in load identification,and the proposed threshold identification method is used to determine the load category under identification.The results show that the identification accuracy of the method in the off-line state is about 1%different from that of k-NN and BP neural network,the identification accuracy of this method in the online state is nearly90%,and it takes less than 10 ms to complete an identification task by means of DSP28335 processor.The two indicators of identification accuracy and identification timeliness prove the effectiveness and feasibility of this method.