Bad data detection enabled estimation based PSS for a grid connected generator

• Suppression of the low frequency oscillations and enhancement of the stability of a grid connected synchronous generator through estimation based PSS (EB-PSS). • The proposed estimation based PSS doesn't need any mechanical measurement and hence doesn't relies on sensors. • The scheme is enabled with the bad-data detection which alarms when the input electrical signals or measurements are incorrect. • The proposed scheme has been tested in SMIB and WSCC 3-machine 9-bus power systems under various disturbances and observed that it significantly improves the dynamic stability without requiring additional sensors. The Power System Stabilizers (PSS) placed in the generators mitigate the low frequency oscillations and ensure the stable operation of the power system. A conventional PSS takes generator speed as input and passes it through the wash-out and lead-lag compensator to further achieve the suitable input to the excitation system. The generator speed used as input for the PSS is typically a mechanical measurement which can sometimes include errors due to inaccuracies in the speed sensor. This may result in the malfunction of the PSS. To address this issue, this article proposes an Estimation-Based PSS (EB-PSS) that utilizes an estimated generator speed instead of the directly measured value. Thereby, the proposed EB-PSS eliminates reliance on the mechanical speed sensor, enhancing its robustness and reliability. The Unscented Kalman Filter (UKF) has been used as estimator which takes the generator voltage and current as measurements along with the system dynamics and provides the estimated speed for further use in PSS. It is enabled with the detection of the bad-data in the measurements which alarms when poor data are being fetch. It was found that the proposed method significantly improves the dynamic stability of both Single Machine Infinite Bus (SMIB) system and WSCC 3-machine 9-bus power systems under various disturbances, without requiring additional sensors. The results also demonstrate that the proposed approach effectively detects bad measurements.