Dynamic state prediction and hierarchical filtering for power system state estimation (Q1119534)

The state variables currently used in state estimation of the electric power systems are bus voltage magnitudes and relative phases. The authors propose another state vector including active and reactive leads, generated powers and the voltage magnitude in generator nodes. That change allows to construct more effective models of the dynamic processes in the power system. (A reader must take into account that the authors use a specific engineering terminology where the term ``dynamics'' refers to the loads only and the generators are considered as ``static'', inertia-free devices). An extended Kalman filter is constructed to estimate the state on the base of observations which can be obtained from some nodes of the system. To simplify the computation the authors suggest a hierarchical organization. The entire system is decomposed into a number of regions (or subsystems) connected by tie-lines. The two ends of each tie-line are called boundary nodes. The hierarchical algorithm includes the following two-level calculation: (i) a dynamic state estimation, performed independently in each subsystem, upon choosing a local phase reference node and using local measurements; (ii) a coordination of the local estimations, so as to estimate the phase angles of the local reference nodes with respect to a general reference. The coordination level must correspond to the subsystems the estimates of the voltages at the boundary nodes. That principle can be generalized to multilevel schemes. The paper presents the approach in detail. The estimator's effectiveness for real time applications is verified by simulations which have been carried out on the standard IEEE II8-node test system. However, the real effectiveness depends on the adequacy of the load prediction models which are used in the Kalman filter.