Studies On Low Frequency Oscillation Analysis And Wide-Area Adaptive Control For Large Scale Power System

With implementation of the "transmission of electricity from the western to the eastern region" strategy, two big and long-chain AC synchronized power grids have been formed in our country according to a series of interconnection projects. Recently, "ultra- high-voltage (UHV) grid as the backbone grid, coordinated development of power grids at all levels" strategical objective has been proposed by State Grid Corporation of China (SGCC). On January 6, 2009, the lOOOkV Jindongnan-Nanyang-Jingmen UHV AC experimental pilot project, which is researched, developed, designed and constructed independently by China, has been formally completed and put into operation. According to planning, "two vertical two horizontal" UHV backbone grid will be constructed and an ultra-large-scale North China-Central China-East China UHV AC synchronized power grid will be formed by the beginning of the "Twelfth Five-Year". By the year around 2020, a strong structure will be formed for UHV power grid of SGCC, which takes the North China, Central China and East China as the core, and links major regional power grids, large coal bases and large hydropower bases and main load centers.During construction of our country's ultra-high voltage (UHV) power grid, both electromagnetic loop circuit and stressed long UHV transmission passage may lead to low frequency power oscillation. Large scale power system has a characteristic of large scale and variable operation condition. Both of them challenge existing low frequency analysis and control methods. Therefore, studies on low frequency analysis and wide-area adaptive control for large power system, which is used to overcome shortcomings of existing analysis and control methods in practical applications, has a great importance in enrichment and extension of power system low frequency oscillation analysis and control theory, prevention blackout and improvement transmission capacity between regional power systems. Based on literature review of studies on mechanism, analysis and control method on small disturbance angle stability, thorough studies and useful discussions on low frequency analysis method and wide-area adaptive control strategies for large power system have been made in this dissertation. The main research work and innovative fruits are as follows.1) With regards to limitation of conventional eigenvalue analysis method holds true only for middle or small scale power system, a practical method for large scale interconnected power system low frequency oscillation analysis is proposed by combing Prony algorithm and sparse eigenvalue technique. Sparse eigenvalue analysis technique, which makes full use of sparsity of augmented state matrix during small disturbance stability analysis for large scale power system, can compute out concerned eigenvalue subset from all eigenvalues and can analysis power systems that have arbitrary scales. Those concerned partial eigenvalues and eigenvectors can be computed out quickly and accurately by taking results of Prony analysis as initial shift points for the algorithms. Thus adverse effects on computation speed and convergence performance for sparse eigenvalue technique due to inappropriate choice of initial parameters will be avoided. Based on the proposed method, two sparse eigenvalue algorithms, i.e., inverse iteration to Rayleigh quotient iteration (II/RQI) and implicitly restarted Arnoldi (IRA) are taken as examples to conduct low frequency oscillation analysis forNortheast China-North China interconnected power system and North China-Central China-East China UHV synchronized power grid. Critical interarea low frequency oscillation modes are obtained and power system stabilizers (PSSs) are installed on generators in Northeast China- North China interconnected power system to improve system damping. Influencing factors on interarea low frequency oscillation modes for North China-Central China- East China UHV synchronized power grid are also thoroughly analyzed.Low frequency oscillation analysis results for Northeast China- North China interconnected power system show that there is a weak damping interarea low frequency oscillation mode where generators in Shandong power grid swing against those in Northeast China power grid. After low frequency oscillation analysis for North China-Central China-East China UHV synchronized power grid, following conclusion is drawn. There are six main interarea low frequency oscillation modes in the system. Damping ratio of the oscillation mode where generators in Shandong power grid swing against generators in Western Inner Mongolia power grid increases along with increment of power transmited from North China main grid to Shandong power grid. It declines sharply after the second interconnection project between North China main grid and Shandong power grid, Huanghua-Binzhou 500kV double-circuit transmission line, is put into operation. Extremely, when North China main grid needs emergency power from Shandong power grid, damping decline of this mode becomes more serious. Increment of power transmited from Western Inner Mongolia power grid to North China main grid helps enhance damping of modes in which generators in Western Inner Mongolia power grid participate. Damping ratio of mode where generators in Fujian power grid swing against generators in East China main grid increases along with increment of power transmited from Fujian power grid to East China main grid. Damping of this mode is notably enhanced after all UHV transmission lines in Hangbei substation are lost, while damping of oscillation mode where generators in Anhui and Zhejiang power grid swing against generators in Jiangsu power grid and Yangcheng power plant declines sharply after the lost. Whether Shijiazhuang UHV substation is lost or not, has little impact on all interarea low frequency oscillation modes.2) In power system damping control research, system model is premise and basis for controller design. With respect to modern interconnected power system, which has a characteristic of large scale and variable and complex operation conditions, online recursive closed-loop subspace identification algorithm, i.e., past output errors in variables-multivariable output error state space model identification algorithm (POEIV-MOESP), is proposed. It is the first time that this algorithm is proposed for damping control. The algorithm mainly focus on solving "identification" in "control while identification" thought. Under closed-loop condition, system input and output measurement noises as well as process noise are eliminated by instrumental variable which is formed by system "past" input and output Hankel matrices. Then, space spanned by the columns of system extended observability matrix is obtained by matrix orthogonal projection. With the help of singular value decomposition (SVD) and shift invariant characteristic of extended observability matrix, system matrices A and C can be derived. Thereafter, matrices B and D can be computed out by using least square algorithm. When new sampled input and output data become available, extended instrumental variable- projection approximate subspace tracking algorithm (EIV-PAST) is used to recursively update A and C, while recursive least square (RLS) algorithm is utilized to recursively update B and D. Aspects on applying the algorithm to power system damping control, such as closed-loop identifiability of system model, choice of persistent excitation signal and scaling factor for sampled data, are also discussed in this dissertation. Under the self-tuning principle based wide-area adaptive damping control framework, linear quadratic optimal partial output feedback supplementary damping controller is designed by using online recursive closed-loop subspace identification algorithm.The algorithm use system dynamic response to identify reduced-order model which contains system dominant low frequency oscillation modes. Thus, difficulty in obtaining practical system model which has very high order is avoided. Additionally, the algorithm has good numerical stability and low time complexity, which provides basis for "control" in "control while identification" thought.Simulation results of the China EPRI 8-machine 36-bus system demonstrate that the proposed on-line recursive closed-loop identification algorithm can effectively identify and track reduced- order state space model which contains system dominant low frequency oscillation modes, and can be used to adjust parameters of supplementary damping controller on-line. Additionally, supplementary damping controller can effectively damp system intearea low frequency oscillations.3) With regards to shortcoming of optimal control which cannot take control constraints into the optimization, wide-area adaptive damping control strategy based on model predictive control (MPC) is proposed, which attempts to solving "control" in "control while identification" thought. Based on identified model, system output disturbance model is obtained by augmenting system states with output disturbances to prevent outputs from steady-state offsets. Kalman filter is used to estimate states the augmented model. According to the identified model, predictive equations for closed-loop system are formulated. Then an objective function representing the cost of deviation of system responses from the reference trajectory and the cost of imposing damping controls in infinite horizon is established. Considering constraints on control input, optimal control can be obtained by solving this optimal control problem which uses current state of power system as the initial state. Online model identification and control optimization are repeated in each time interval. The proposed method is an organic integration of online model identification and online updating control parameters. It accomplishes adaptive control of power system low frequency oscillation. Thus, the inherent shortcomings of controllers with fixed parameters based on offline identification are overcome and the problem of the control performance degradation due to variation of the complex operation conditions and time-varying and uncertain characteristic of system parameters are solved. Negative effects of control are avoided by adopting model prediction, receding optimization and feedback compensation strategy in this method.Simulation results of the China EPRI 8-machine 36-bus system demonstrate that time consumption for updating of system state space model and model predictive control law is less than sampling interval. Thus, model predictive damping controllers satisfy requirements for online application. It can effectively damp inter-area low frequency oscillation modes. It also have the ability to coordinate with PSSs and other model predictive damping controllers in multi-machine power systems and the ability to adapt to changes in operation conditions.