State Estimation

The module runs the non-linear and DC state estimation, as well as the linear state estimation with PMUs only. By default settings, the state estimation algorithms use the weighted least-squares estimation, but it is also possible to use the least absolute value estimation [1, Sec. 6.5].

The non-linear state estimation is implemented using the following features:

• solved by the Gauss-Newton method,
• the state vector is given in the polar coordinate system,
• phasor measurements are given in the polar coordinate system,
• measurement errors are uncorrelated.

The linear state estimation with PMUs only is implemented using the following features:

• the state vector is given in the rectangular coordinate system,
• phasor measurements are transformed from polar to rectangular coordinates,
• the covariance matrix is transformed from polar to rectangular coordinates.

Besides state estimation algorithms, we have implemented the bad data processing using the largest normalized residual test [1, Sec. 5.7]. The routine proceeds with bad data analysis after the estimation process is finished, in the repetitive process of identifying and eliminating bad data measurements one after another.

Finally, to achieve global observability of the power system only with PMUs, we implemented the optimal placement algorithm given in [2].

References

[1] A. Abur and A. Exposito, “Power System State Estimation: Theory and Implementation,” ser. Power Engineering. Taylor & Francis, 2004.

[2] B. Gou, “Optimal placement of PMUs by integer linear programming,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 1525–1526, Aug. 2008.