TY - JOUR
T1 - AI-Based Damping of Electromechanical Oscillations by Using Grid-Connected Converter
AU - Baltas, Gregory N.
AU - Lai, Ngoc Bai
AU - Tarraso, Andres
AU - Marin, Leonardo
AU - Blaabjerg, Frede
AU - Rodriguez, Pedro
PY - 2021/3
Y1 - 2021/3
N2 - The proliferation of grid-connected converter interfaced energy sources in Smart Grids, enhance sustainability and efficiency as well as minimizing power losses and costs. However, concerns arise regarding the stability and reliability of future smart grids due to this wide integration of power electronic devices, which are recognized to affect the dynamic response of the system, especially during disturbances. For instance, apart from the lower damping of existing electromechanical modes, new low-frequency oscillations begin to appear. Yet, the ability of grid-connected converters to provide grid support functionalities can alleviate the aforementioned challenges. Relevant studies show that these functionalities can be enhanced even further, if information regarding the oscillation characteristics are available. Traditional methods for extracting modal information are very well suited for monitoring purposes, however, they pose certain limitations when considered for control applications. Therefore, this paper proposes a multi-band intelligent power oscillation damper (MiPOD) that exploits 1) the inherent characteristics of grid-connected converters to damp multiple power oscillations and 2) the modeling capabilities of Artificial Intelligence (AI) for predicting the frequency of electromechanical oscillations in the system, as operating conditions change. Essentially, the MiPOD integrates the AI model in the control loop of the converter to attenuate multiple modes of oscillation. The proposed controller is validated for different disturbances and randomly generated operating points in the two area system. Specifically, in this case the AI model is a Random Forest ensemble regressor that is developed for tracking two electromechanical modes. As it is shown, the MiPOD can improve the overall performance of the system under various contingency scenarios with only 6% of the corresponding total nominal capacity of synchronous generators. In addition, the monitoring and damping abilities of the MiPOD are demonstrated for a vast range of operating points just by tuning two parameters; the predicted oscillation frequencies of the local and inter-area mode.
AB - The proliferation of grid-connected converter interfaced energy sources in Smart Grids, enhance sustainability and efficiency as well as minimizing power losses and costs. However, concerns arise regarding the stability and reliability of future smart grids due to this wide integration of power electronic devices, which are recognized to affect the dynamic response of the system, especially during disturbances. For instance, apart from the lower damping of existing electromechanical modes, new low-frequency oscillations begin to appear. Yet, the ability of grid-connected converters to provide grid support functionalities can alleviate the aforementioned challenges. Relevant studies show that these functionalities can be enhanced even further, if information regarding the oscillation characteristics are available. Traditional methods for extracting modal information are very well suited for monitoring purposes, however, they pose certain limitations when considered for control applications. Therefore, this paper proposes a multi-band intelligent power oscillation damper (MiPOD) that exploits 1) the inherent characteristics of grid-connected converters to damp multiple power oscillations and 2) the modeling capabilities of Artificial Intelligence (AI) for predicting the frequency of electromechanical oscillations in the system, as operating conditions change. Essentially, the MiPOD integrates the AI model in the control loop of the converter to attenuate multiple modes of oscillation. The proposed controller is validated for different disturbances and randomly generated operating points in the two area system. Specifically, in this case the AI model is a Random Forest ensemble regressor that is developed for tracking two electromechanical modes. As it is shown, the MiPOD can improve the overall performance of the system under various contingency scenarios with only 6% of the corresponding total nominal capacity of synchronous generators. In addition, the monitoring and damping abilities of the MiPOD are demonstrated for a vast range of operating points just by tuning two parameters; the predicted oscillation frequencies of the local and inter-area mode.
KW - Artificial Intelligence
KW - electromechanical oscillations
KW - power electronics
KW - random forests
KW - grid-connected converter
UR - http://www.scopus.com/inward/record.url?scp=85102934479&partnerID=8YFLogxK
U2 - 10.3389/fenrg.2021.598436
DO - 10.3389/fenrg.2021.598436
M3 - Journal article
SN - 2296-598X
VL - 9
SP - 1
EP - 17
JO - Frontiers in Energy Research
JF - Frontiers in Energy Research
M1 - 598436
ER -