Research On Sub-synchronous Oscillation Detection Method Based On Broadband Measurement

With the rapid development of the power industry,a large number of new energy sources such as wind power and photovoltaic are connected to the power grid,resulting in frequent occurrence of sub-synchronous oscillations,which seriously affects the safe and stable operation of the power system,and broadband measurement devices and PMUs are continuously installed in the power grid,providing conditions for online fast identification of sub-synchronous oscillations.However,the existing monitoring system can hardly meet the requirements of monitoring sub-synchronous oscillations,and there is a lack of targeted device optimization configuration for sub-synchronous oscillations;at the same time,the current PMU-based sub-synchronous oscillation identification method also has problems such as large calculation volume of frequency domain analysis and troublesome feature extraction.Therefore,the thesis focuses on three aspects:optimal configuration of monitoring devices,feature extraction of broadband measurement data and sub-synchronous oscillation identification methods.The main research work and innovative results done in this thesis are as follows:To address the need for optimal configuration of monitoring devices to effectively monitor sub-synchronous oscillation-prone nodes,an optimal configuration of sub-synchronous oscillation monitoring devices for wind power grid-connected systems is proposed using the property that the aggregated short-circuit ratio reflects the magnitude of sub-synchronous oscillation risk.The impact of line N-1 faults and zero-injection nodes are further considered and validated using the improved IEEE14 and IEEE39 node standard models as well as the actual model in Northwest China to ensure certain reliability and economy while achieving focused monitoring of sub-synchronous oscillations.To address the problem of how to effectively extract the features reflecting subsynchronous oscillations from a large amount of complex data,a large amount of actual subsynchronous oscillation event data is analyzed,and two oscillation dispersion scenarios existing at the time of the accident are defined;a linear mapping-filled feature characterization method is proposed to convert the one-dimensional feature data into two-dimensional features for the characteristics of whether the two data are sorted smoothly as well as continuously changing at oscillation and normal images.The proposed features can cope with the effects of different noise levels and the number of oscillation components,and have the advantage of being simple and easy to distinguish.To address the problem of how to use the proposed feature images to achieve fast recognition of subsynchronous oscillations in a variety of complex scenes,a CNN image recognition-based method for recognizing subsynchronous oscillations with short window length is proposed,and the most accurate and stable recognition model for both types of data is established by optimally adjusting the CNN parameters and testing and comparing the performance under different feature extraction data window lengths.The most accurate and stable recognition model is established for both types of data,and the more accurate recognition results are obtained by combining the results of both,and the fast and accurate recognition of subsynchronous oscillations is achieved.