Hybrid method for power system state estimation

Abstract

State estimation in power systems is classically based on the weighted least squares method. Recently, different extensions of Kalman filters have been proposed. Among them, the 'unscented' Kalman filter (UKF) improves the results of weighted least squares methods, when there are small changes in the system, as it considers the history of the state. The novel algorithm presented in this work combines the best of both approaches. To perform this task a new index is defined to allow the algorithm to choose in real time, and for each iteration, between a static or a dynamic estimator. This combination allows overcoming the anomalies observed when the UKF faces abrupt variations of the system state and also the lack of observability that weighted least squares could present. The proposed methodology was tested with three test cases outperforming the previously mentioned algorithms.

Recent advances on state estimation for power grids with unconventional measurements

“…Different from the EKF, the Kalman filter has the desirable properties of convergence in estimation error and low computational complexity. Though the SSE and DSE methods are summarised separately above, a hybrid filter combining the static WLS and the dynamic UKF has been developed to exploit the advantages of both methods [55].…”

Neural network-based power system dynamic state estimation using hybrid data from SCADA and phasor measurement units

“…On the other hand, load forecasting is also applied by some researchers as a method to derive state transition model, i.e., process model . As dynamics of power system is mainly driven by loads and generation, the latter approach seems to be more realistic than smoothing techniques . Loads are usually independent from each other, and it means they can be considered as uncorrelated variables.…”

“…Besides applying UKF, nodal power injections from both load points and DG units are projected and transformed into state predictions through load flow computation as an approach to derive state transition model . To enhance the numerical stability of the UKF used for power system DSE, a new UKF method with guaranteed positive semidefinite error covariance matrix is proposed and compared with some existing similar approaches in Qi et al A hybrid method is proposed by Risso et al, with the aim of taking advantages of both WLS and UKF simultaneously. A GPU‐based 2‐level dynamic state estimator is proposed in Karimipour and Dinavahi that is based on the EKF utilizing both SCADA and PMU measurements.…”

“…26 As dynamics of power system is mainly driven by loads and generation, the latter approach seems to be more realistic than smoothing techniques. 23,27 Loads are usually independent from each other, and it means they can be considered as uncorrelated variables. While nodal loads are predicted, the state transition model can be derived based on load flow calculations.…”

Parallel relaxation‐based joint dynamic state estimation of large‐scale power systems

“…Conventional methods to improve the cycle time for state estimation include the following main strategies: † Complexity reduction to alleviate the computational burden using reduced order model, lower dimensional measurement data or partial update of the Jacobian matrix [5][6][7]. † Hierarchical two-level state estimation which decomposes the whole system into several independent subsystems wherein each subsystem uses its own measurement set to estimate the states locally and then sends the estimated states to a centralised coordinator [8][9][10]. † Distributed state estimation where each subsystem can run its local state estimation and exchange data between the substations without the central coordinator [11][12][13][14].…”