基于虚拟通道延拓的局部放电信号压缩采样方法

Compressed Sampling of Partial Discharge Signal Based on Virtual Channel Extension

针对局部放电信号频域非带限且要求采样频率过高而引发的电路参数与性能矛盾,提出一种基于虚拟通道延拓的局部放电信号压缩采样方法,利用压缩感知并行测量来降低采样频率。通过在采样电路中实现虚拟通道延拓,降低了并行测量通道数,优化了压缩采样电路结构。为了使测量通道参数适合局部放电信号特征,提出了基于非相干性延拓的三值观测矩阵设计算法,包括引入自相干正则项和利用时序加窗差值的变步长梯度下降法来降低参数相干性。最后,将观测矩阵用于虚拟通道延拓电路,实现局部放电信号的压缩采样。结果表明,文中的压缩采样方法将采样频率降低至1.6 MHz,利用多通道测量将整体采样频率和数据量分别压缩至48%和14%,将实际并行测量通道数优化至12条。该文的压缩采样方法降低了传感设备的元器件参数要求和电路结构复杂度,达到了电力设备局部放电信号在线监测的需求。

Partial discharge(PD)is the important parameter to monitor the operating status of power equipment.However,PD signals are non-band-limited in the frequency domain,and the sensor theoretically needs to be equipped with the analog-to-digital converter(ADC)with infinitely high sampling frequency.The cost of ADCs increases exponentially with the sampling frequency due to production process limitations.Aiming at this circuit parameter and performance contradiction,the compressive sensing(CS)theory is introduced to achieve compressed sampling of PD signals from power equipment.A compressed sampling method for PD signals based on virtual channel delay topology is proposed to reduce the sampling frequency by using compressed sensing parallel measurement.Firstly,Analog-to-information converter for PD signal based on virtual channel extension(VCE-AIC)is proposed in combination with PD signals characteristics.Parallel measurements are used to reduce the sampling frequency and to create virtual measurement channels based on non-coherent extension of submeasurement segments.The actual number of measurement channels is reduced,and the compressed sampling circuit structure is optimized.After that,according to the characteristics of PD signals,channel delay requirements and noncoherence criterion,a three-valued observation matrix design algorithm based on noncoherent delay is proposed.It includes the introduction of a self-consistent regular term and a variable-step gradient descent method using the time-series plus window difference to reduce the parameter coherence.Finally,the observation matrix is used in the virtual channel delay topology circuit to realize the compressed sampling of PD signals.The reconstruction accuracy of the compressed measurement signal is guaranteed from the following six perspectives.(Ⅰ)The observation matrix is designed to satisfy the RIP criterion,i.e.,the observation matrix has low coherence with the sparse dictionary.(Ⅱ)The proposed three-valued observation matrix design algorithm based on non-coherent extension for VCE-AIC.(Ⅲ)Maximize the completeness and generality of the sparse dictionary for PD signals.(Ⅳ)Adopting orthogonal matching tracking algorithm for signal reconstruction.(Ⅴ)The generally common PD signals waveform evaluation index is selected.(Ⅵ)The complete theoretical and experimental analysis is set up for the selection of each parameter.Sampling analysis is performed based on the measured PD signals.The results show that the compressed sampling method reduces the sampling frequency to 1.6 MHz.The overall sampling frequency and data volume are compressed to 48%and 14%,respectively,using multi-channel measurements.The actual number of parallel measurement channels is optimized to 12.The following conclusions can be drawn from the analysis of the experimental results.(Ⅰ)VCE-AIC reduces the ADC sampling frequency and optimizes the circuit structure.VCE-AIC uses multi-channel measurement to compress the overall sampling frequency to 48%and optimizes the actual number of parallel measurement channels to 12.VCE-AIC significantly improves the circuit implementability with an easy-to-implement three-value observation matrix.(Ⅱ)VCE-AIC can compress the data volume up to 14%,which greatly reduces the data communication volume of PD monitoring.(Ⅲ)VCE-AIC effectively solves the circuit and performance conflicts faced by online monitoring of PD signals due to its multi-channel compressed sampling and circuit structure optimization.In addition,it also has data security and limited denoising capability.The compressed sampling method in this paper reduces the component parameter requirements and circuit structure complexity of the sensing equipment and meets the demand for online monitoring of PD signals of power equipment.