Dynamic State Estimation of Microgrid With Imperfect Data Communication
AuthorRahaman, Meh Fuzur
AdvisorFadali, M. Sami
Electrical and Biomedical Engineering
StatisticsView Usage Statistics
Dynamic state estimation of power systems is essential for wide area control purposes. In this thesis, we present the results of dynamic state estimation for a grid-connected microgrid including two synchronous generators and three loads. The Unscented Kalman filter (UKF) and the Extended Kalman filter (EKF) are implemented using a classical generator model connected to a Thevenin equivalent of the remainder of the microgrid. The model is used to estimate the six states variables of the generator; namely, rotor angle, speed variant, d- and q- axis transient voltages, d-axis damper flux, and q-axis second damper flux. Both real power and reactive power are used as measurements in our state estimation algorithm. The estimation results are compared with the true values to demonstrate the accuracy of the state estimator. In addition to data loss or delay, sensor measurements may include outliers that distort state estimation. We utilized the Generalized Maximum Likelihood-extended Kalman filter (GM-EKF), as a robust estimator, which exhibits good tracking capabilities suppressing the effects of bad data (outliers). We also used two methods of state estimation on UKF to deal with bad data. Simulation results obtained from the UKFs are compared with those of GM-EKF. We present simulation results at a high frequency of 1 kHz of state estimation for different scenarios that include normal operation, fault at Point of Common Coupling (PCC), loss of generator, and loss of load. We also developed a scheme to use delayed data in Kalman filter estimation and used it to simulate the effect of data loss and/or delay in the communication system of the microgrid. For the same scenarios, we also present simulation results at 50 Hz, which is compatible with Phasor Measurement Units (PMU), including bad data as well as data loss or delay. Our results demonstrate that while both filters successfully detect bad data, the UKF methods provide better estimates than those of the GM-EKF.