Research On The Key Technologies Of Online Analysis Of Power System Low Frequency Oscillation

With the development of our national economy, energy consumption and grid scale continue to grow. This makes the stability problems in power system become more complex and its analysis and control are getting harder. In fact, the low frequency oscillation in power system which seriously affects the stable operation of the power system has repeatedly occurred and become one of the key issues which threaten the safety and stability of interconnected power grid and restrict the power transmission capacity.Traditional methods are based on the power system structure model, such as applying the small signal stability theory to analyze low frequency oscillation modes or Normal Form and modal series method with the non-linear higher order terms considered to analyze low frequency oscillation modes. The analysis accuracy of these methods is subject to the order of the model, the parameters accuracy, un-modeled dynamic and interactive modal factors, so that traditional low frequency oscillation analysis methods have been unable to meet the requirements of stability analysis of modern power system. Nowadays, with the extensive application of WAMS (Wide-Area Measurement System) in the wide area power grid, the analysis of low frequency oscillation of power system based on the measurement signals becomes a hot research topic.With measured low frequency oscillation signals, the disturbed trajectories reflect the true dynamic process of the system and contain the complete characteristic information of the excitation modes. However, the low frequency oscillation signal is a typical non-stationary signal, which demonstrates non-stationary signal characteristics such as the amplitude changing over time and the random emergence of excitation modes. Traditional measurement signal based analysis methods assume the signal is stable, which limits its use to analyze non-stationary characteristics of the low frequency oscillation modes. Therefore, the data-driven HHT (Hilbert Huang Transform) method is proposed to analyze the low frequency oscillation. In this paper, the main theories and key technical issues of HHT method are studied as following: 1. Low frequency oscillation analysis based on the measured signals. Comparative study is performed on theoretical models and algorithms process of the improved Prony method and the HHT method analysis of low frequency oscillation signal. The limitation of Prony analysis to analyze low frequency oscillation signal is tested and discussed, such as model order, fitting precision and non-stationary parameters identification. Using the HHT method to deal with non-stationary oscillation signal and identify oscillation mode characteristic parameters is studied in detail, including discussing the physical meaning of the method to identify the low frequency oscillation mode and analyzing the results validity and accuracy of low-frequency oscillation mode parameters identification.2. Analyze the main factors affecting the validity and accuracy of the low frequency oscillation signal based on HHT method, then some key technologies and improved methods to improve the performance of HHT method are studied in detail. For the problem of end effect, fitting error, mode mixing and damping loss existed in HHT method, the low-pass digital filtering technology, the endpoint extension technology based on AR model, the B-spline interpolation technologies, the sifting control factors and complex wavelet analysis method are proposed to achieve an improved EMD (Empirical Mode Decomposition) process, which greatly improves the accuracy of low frequency oscillation mode characteristic parameters identification.3. Especially in terms of real-time measurement oscillation signals provided by WAMS, the online low frequency oscillation analysis techniques based on HHT is studied in detail. As the EMD process has mode mixing and non-stationary parameter identification problems in low frequency oscillation signal, improved frequency heterodyne method and adaptive sliding window technologies are proposed to deal with the intermittent and octave mode mixing problem within EMD decomposition process. The proposed method is able to determine the presence time of different low frequency oscillation modes and implement the Is hexadecimal near-real-time analysis and identification algorithm of low frequency oscillation signal. The algorithm can adaptively implement the online analysis of measured signal of low frequency oscillation, which has high accuracy and instantaneity to identify the low frequency oscillation mode parameters.4. The limitations of small signal stability analysis theory to analyze the low frequency oscillation modes correlation with the structure-based model is discussed in detail. The contribution factors, considered from the perspective of energy conversion are established based on the HHT method and measured oscillation signals, which can define the correlation between the disturbance and the low frequency oscillation mode quantitatively. Furthermore, an online algorithm to compute contribution factors based on the HHT method is proposed to solve the correlation problem between the source generators and the oscillation mode when weakly damped oscillation mode and forced oscillation mode appear. The proposed method can further consolidate the theoretical base of low frequency oscillation analysis based on the HHT method and expand its applications.