基于对角化LDPC压缩感知和k-近邻算法的广域系统宽频振荡监测方法

Monitoring Method of Wide-band Oscillation Based on Compressed Sensing of Diagonalizable LDPC and k-nearest Neighbour Algorithm

在“双高”电力系统中,宽频振荡的发生概率大大增加。然而,传统基于广域测量系统(wide area measurement system,WAMS)的振荡监测方法一方面监测的振荡频带范围过窄,另一方面其准确性和快速性难以适应复杂的电网运行状态。因此提出一种基于对角化低密度奇偶校验码(low-density parity-check codes,LDPC)校验矩阵和k-近邻算法(k-nearest neighbor,KNN)的宽频振荡监测方法。首先,基于对角化LDPC校验矩阵对电力系统信号进行压缩采样,大大减少了宽频振荡的数据传输量,有利于在现有相量测量单元(phasor measurement unit,PMU)上传频率下实现几百Hz的宽频振荡数据的传输。在此基础上,主站直接基于压缩采样值作为输入特征,采用加权KNN算法进行振荡检测,避免了人为设置阈值带来的误判,提高了振荡检测的快速性和准确性。最后,根据振荡检测结果,采用正交匹配追踪(orthogonal matching pursuit,OMP)算法,在主站准确重构宽频振荡信号,便于广域系统的振荡全局性分析。仿真结果表明所提方法在噪声、数据缺失和数据有误等情况下,仍然能够实现宽频振荡信号的快速准确监测。

In the“double high”power system,the probability of wide-band oscillations is greatly increased.However,it is difficult for the traditional methods based on the wide area measurement system(WAMS)to adapt to the complex power grid due to their narrow ranges of the oscillation frequency and their restricted accuracy and rapidity.Therefore,a wide-band oscillations monitoring method based on the diagonalized low-density parity-check codes(LDPC)check matrix and the k-nearest neighbor(KNN)algorithm is proposed.Firstly,the power system signals are compressed based on the check matrix to reduce the transmission data,which is beneficial to achieve hundreds of Hz wide-band oscillation data transmission under the existing upload frequency of phasor measurement unit(PMU).On this basis,the master station uses the weighted KNN to detect the oscillation directly based on the compressed value as the input feature,avoiding the false judgment caused by the artificial setting threshold and improving the rapidity and accuracy of the detection.Finally,according to the detection results,the orthogonal matching pursuit(OMP)algorithm is used to accurately reconstruct the signals at the master station,which is convenient for the global analysis of the wide area system oscillations.The simulation results show that the proposed method can realize the fast and accurate monitoring of the wide-band oscillation in the case of noise,data missing and data error.