An Approach to Suppress Low-Frequency Oscillation by Combining Extended State Observer With Model Predictive Control of EMUs Rectifier

Recently, low-frequency oscillation (LFO) occurs in electrified railways, which exhibits a synchronous periodic oscillation of about 2–7 Hz in the traction network and electric multiple units (EMUs). To suppress the LFO, an improved model predictive control (MPC) combining with extended state observer (ESO) is proposed. First, for the design of ESO, the changes of system parameters and the unmodeled item in vehicle mathematical model are extended to a new variable and then estimated to compensate the calculated control voltage in real time. Second, the observer stability is analyzed by obtaining the transfer function, and the appropriate gains of observer are chosen through the pole assignment technique. At last, by the combination of the predictive currents, the optimal control voltages can be obtained. The control performance among MPC,
dq
decoupling control, passivity-based control, and the proposed method is compared. In addition, the system robustness when system parameters change is discussed. At last, to further verify the effectiveness of the proposed method for suppressing LFO, an integrated dSPACE semi-physical experimental platform including eight simulated EMUs and an equivalent traction network is constructed. Simulation and experimental results show that the proposed method not only effectively suppresses LFO but also accelerates the response speed of traction line-side converter, reduces the distortion of grid-side and vehicle-side currents, and overcomes the disadvantages of MPC and
dq
decoupling control under the mismatching of system parameters.