The Study Of Distributed Power System State Estimation

Since the network structure and operation modes of the power system is becoming complex, the modern scheduling system is being required to grasp the actual operational states of the power system more quickly, accurately and comprehensively. In order to ensure the power system to operate more safely and economically, the modern scheduling system also need to make a decision according to the different operational problems. State estiamtion, also known as the filtering, can estimate or predict the system operational states, which provides the real-time reliable data for the power system security assessment, economic dispatch and online control. At present, the centralized state estimation has been implemented in the actual power networks, which needs to collect all real-time measurements of the power system and maitain the whole network model. With the system dimension growing, the central computer at the dispatch center will undertake the heavy tasks of the data communication, storage and processing. Moreover, with the number of PMUs increasing, the big data problem will have to be faced after WAMS becoming mainstream. Therefore, this thesis focuses on how to deal with the problem of distributed state estimation in large-scale power system. The main contents are as follows:Most existing decentralized state estimation methods have a slow convergence speed or some specific requirements on the comunication topology. To overcome above disadvantages, this thesis proposes a new decentralized state estimation algorithm, which is applicable to the normal measurement case, based on the muli-area constrained weighted least square estimation model and the finite-time average concensus protocol. The proposed method has very fast convergence speed, also can adapt to any communication topology by simple parameter settings. Moreover, by considering the presence of abnormal measurements, we deduce a modified algorithm according to the geometric meaning of the least squares estimate. After eliminating the abnormal measurements in each sub-region, the modified algorithm can converge to the centralized estimated values without changing the information matrix, by executing the finite-time average consensus protocol several times.A new hierarchical dynamic state estimation method based on Cubature Kalman Filter is proposed with the coordination at the state estimation level, which can deal with the difficulties in the most existing studies such as the internal states coordination and the comunicaiton problem. By using CKF in the local estimation, the centralized boundary states coordination method is given using the partitial least squareestimation fusion technique. According to the boundary states coordinated values, the internal states of the local estimation can be corrected by the linear equality constrainted Kalman Filter. The test simulation results show that the estimated precision of the proposed method is close to the centralized method, and the proposed method has better real-time property. Moreover, the proposed method only need very few communication data, which also can be easy to be implemented.The most existing hierarchical static state estimation methods with the coordination at the state estimation level are without internal states coordination and need iterative caculations in boudary-states coordination. For dealing with above disadvantages, a new hierarchical static state estimation method based on the covariance matrix transformation and the estimate projection is proposed. In order to adapt to the case of the measurement system with PMUs or without PMUs, the method for processing the local estimate results before and behind the boudary states coordination is proposed by the linear transformation theory of Gaussian random variables. In the coordination center, the local least square estimation fusion also be used to coordinate the boundary states, obviously reducing the computation of the coordination center. The local estimation of internal states in each area is corrected by the estimate projection method, so the local estimated precision can be improved effectively.Aiming at the effects of model parameter deviations in the synchronous generator dynamic state estimation, the proposed estimation method considers the parameter uncertainties of the synchronous generator model. First, the synchronous generator state estimation model with uncertain parameters is established. Then a ellipoid-based set membership filter for the synchronous generator dynamic state estimation is proposed by the set operations and interval analysis. In the system disturbance simulation experiment, the state tracking performance and the real-time performance of the proposed method are tested and verified.