Distributed Dynamic State Estimation Algorithms And Applications In Power Systems

Power system state estimation is a vital part of the energy management system,the estimation precision and reliability of which directly influence the accuracy of power system dispatch,security analysis,and operation and control actions.Dy-namic state estimation not only can estimate the system state with better accuracy,but also has the forecasting ability,which is not qualified by the static state estima-tion.The ability of one-step-ahead state prediction provides prior information and more operating time for on-line functions of power systems,such as economic dis?patch and preventive control.So it is significant to the practical applications,and has been widely investigated.With the growing size and the increasingly strengthened interconnection of power systems,and the rapid development and widespread use of measuring and processing device(PMU),which has higher precision and faster up-dating rate of measurements,the traditional centralized state estimation methods are difficult to satisfy the accuracy and real-time demand.Therefore,it promotes the de-velopment of distributed state estimation algorithms for power systems.By dividing the considered system into several nonoverlapping sub-areas,this thesis is devoted to the research on the distributed dynamic state estimation for power systems.The proposed algorithms are scalable to large-scale power systems,which are significant to the theoretical research and practical applications.The main contributions and innovations of this thesis are as follows:(1)Through the domain decomposition of power systems and based on SCADA and PMU hybrid measurements,the distributed dynamic state estimation algorithms are proposed,which make the computation being performed independently and in parallel for d-ifferent subsystems,and increase the total computation speed.(2)Under very few measurement data and the information transmitted from neighbors,but no need to know the globally topological observability,each subsystem use the proposed al-gorithm to estimate its local state,which significantly reduces the computational complexity of the overall process.The proposed algorithms have no central con-trol center,which eliminates the bottleneck problem of the centralized methods in the process of data transferring,and facilitates the implementation and management.(3)The estimation precision of the proposed algorithms is slightly worse than the centralized method,but more accurate than the distributed static methods.When the power system encounters anomaly conditions,the favorable performance,the robustness of the distributed algorithm,and the effectiveness of parameter identifi-cation are verified.(4)Under the relaxed constraint condition,we prove that the error covariance matrices of state estimation and prediction obtained by the proposed al-gorithm are respectively positive definite and have upper bounds,which verifies the feasibility of this algorithm.In the order of chapters,the main contents and results are listed as follows:1.Consider the distributed dynamic state estimation problem for discrete-time linear power systems.Based on the maximum a posteriori estimation technique,a fully distributed state estimation algorithm is proposed,in which the measurement data is obtained by the hybrid measurement system consisting of SCADA and PMU.Firstly,partitioning the considered power system into some nonoverlapping sub-areas,each corresponding subsystem estimates its local state,instead of the whole system state,by using the prior information,local and edge measurements as well as the information transmitted from the neighboring subsystems.Compared with the centralized methods,the proposed algorithm can effectively reduce the dimension of state vector and the computational complexity for each subsystem.Secondly,when the network formed by the partitioned subsystems is acyclic,it is proved that the local state estimate of each subsystem obtained by the distributed MAP estimator converges after a finite number of iterations at each time instant.At last,simulation results verify the efficiency and feasibility of the proposed algorithm for large-scale power system state estimation.2.Based on the extended Kalman filter technique,the distributed dynamic s-tate estimation algorithm for the nonlinear power systems is obtained.When the local measurement of some subsystem is waived for some reasons,it is proved that the proposed algorithm is still applicable,and using the edge measurements and the exchanged boundary information with neighboring areas,each subsystem can also obtain the desired local state estimate.The identification of model parameters on-line improves the accuracy of state prediction and enhances the robustness of the proposed algorithm.When the considered power system encounters abnormal con-ditions,including sudden load change,presence of bad data and topology error con-dition,simulation results in detail confirm the robustness,probably correct estimate values and the effectiveness of the identified parameters for state estimation.Perfor-mance comparison with some known algorithms shows the merits of the proposed algorithm in applications.3.The distributed state estimation problem for nonlinear dynamic systems is further studied.The constraint condition required by the proposed algorithm is re-laxed,and the boundedness of the proposed algorithm is analysed here.The mathe-matical induction is used to prove that the error covariances of state estimation and prediction obtained by the distributed algorithm are respectively positive definite,and the rank criterion of the observability for time-varying system proves that the error covariance matrices have upper bounds,which verifies the feasibility of the proposed algorithm.A new method for parameter identification effectively restrains the detrimental effect caused by sudden load change on the estimation precision.