A robust extended Kalman filter for power system dynamic state estimation using PMU measurements

Marcos Netto,Junbo Zhao,Lamine Mili

DOI: https://doi.org/10.1109/PESGM.2016.7741374

2021-04-06

Abstract:This paper develops a robust extended Kalman filter to estimate the rotor angles and the rotor speeds of synchronous generators of a multimachine power system. Using a batch-mode regression form, the filter processes together predicted state vector and PMU measurements to track the system dynamics faster than the standard extended Kalman filter. Our proposed filter is based on a robust GM-estimator that bounds the influence of vertical outliers and bad leverage points, which are identified by means of the projection statistics. Good statistical efficiency under the Gaussian distribution assumption of the process and the observation noise is achieved thanks to the use of the Huber cost function, which is minimized via the iteratively reweighted least squares algorithm. The asymptotic covariance matrix of the state estimation error vector is derived via the covariance matrix of the total influence function of the GM-estimator. Simulations carried out on the IEEE 39-bus test system reveal that our robust extended Kalman filter exhibits good tracking capabilities under Gaussian process and observation noise while suppressing observation outliers, even in position of leverage. These good performances are obtained only under the validity of the linear approximation of the power system model.

Systems and Control

What problem does this paper attempt to address?

This paper aims to solve two main problems in power system dynamic state estimation:

1. **Improve the robustness of the state estimator to gross measurement errors (i.e., outliers)**: The traditional Extended Kalman Filter (EKF) is prone to produce offsets when dealing with data containing outliers, resulting in inaccurate state estimation. The paper proposes a Robust Extended Kalman Filter (GM - EKF) based on the Generalized Maximum Likelihood Estimator (GM - estimator), which can effectively suppress the influence of vertical outliers and bad leverage points.

2. **Improve the statistical efficiency of the state estimator under Gaussian noise**: By using the Huber cost function, GM - EKF can achieve high statistical efficiency under the assumption that the process noise and the observation noise follow a Gaussian distribution. The Huber cost function behaves as a quadratic function for small residuals and as a linear function for large residuals, thereby improving the fitting ability for normal data while maintaining robustness.

Specifically, the paper proposes the following methods to solve these problems:

- **Batch processing regression form**: By processing the predicted state vector and PMU measurements together, the tracking speed of the system dynamics is accelerated.
- **Outlier identification**: Use the projection statistics algorithm to identify and suppress outliers.
- **Robust pre - whitening**: By pre - whitening the data, the robustness of the filter is further improved.
- **Robust filtering and solution**: Use the Iteratively Reweighted Least Squares (IRLS) algorithm to minimize the objective function, thereby obtaining more accurate state estimates.
- **Covariance matrix update**: Derive the update formula for the state - estimation - error covariance matrix based on the total influence function to ensure the prediction accuracy of the filter in the next time step.

Through simulations on the IEEE 39 - bus test system, the paper verifies that the proposed GM - EKF has good tracking ability and outlier suppression ability under Gaussian noise. However, the paper also points out some limitations of GM - EKF, such as poor performance in strongly non - linear systems and poor performance under non - Gaussian noise conditions. These limitations provide directions for future research.