Data-Driven Model Predictive Control Method for Wind Farms to Provide Frequency Support

Zizhen Guo,Wenchuan Wu
DOI: https://doi.org/10.1109/tec.2021.3125369
IF: 4.8767
2022-01-01
IEEE Transactions on Energy Conversion
Abstract:As the wind power penetration increases, wind farms are required by the grid codes to provide frequency regulation services. This article develops a fully data-driven model predictive control (DMPC) scheme for the wind farm to provide temporal frequency support. The main technical challenge is the complexity and the nonlinearity of wind turbine dynamics that make the DMPC intractable. Based on Koopman operator (KO) theory, a specialized dynamic mode decomposition (SDMD) algorithm is proposed, which fits a global linear dynamic model of the wind turbines. The performance of learning dynamics is powered through integrating the physical knowledge of the wind turbine into the specialized observables of KO. To stabilize the rotor speeds in frequency regulation, the active power contribution is optimally dispatched in a moving horizon fashion. Simulation results show that the DMPC can efficiently learn and predict the wind turbine dynamics. During the frequency response process, the proposed method can effectively track the frequency support order specified by the utility grid operator while significantly stabilizing the rotor speeds.
What problem does this paper attempt to address?