Abstract
Aiming at the problem of motor speed decrease in direct-drive permanent magnet synchronous generator (D-PMSG) wind power generation system after permanent magnet (PM) demagnetization faults, a demagnetization fault-tolerant control strategy in D-PMSG wind power generation system is proposed. Firstly, the D-PMSG mathematical model is described in normal and demagnetization. Secondly, an extended Kalman filter (EKF) observer is designed to observe the PM flux online. Then, flux linkage parameters are introduced into the two-vector model predictive fault-tolerant control so that the increase of stator current is controlled within the limit range. Meanwhile, the motor speed can follow the change of the given speed. In addition, the improved Luenberger mechanical torque observer is designed in the speed outer ring to deal with the vibration caused by unstable wind speed. Finally, compared with the dual-closed-loop Proportional Integral (PI) control, the experimental results show that the demagnetization fault-tolerant control strategy has smaller speed overshoot and smaller speed fluctuation when the mechanical torque changes. The method can maintain the speed balance when the PM demagnetization faults occur and have stronger fault tolerance and anti-interference ability.
Introduction
Direct-drive permanent magnet synchronous generators (D-PMSG) are widely used in wind power generation due to their high efficiency and lack of external excitation requirements. However, these systems often operate in harsh environments where high temperatures and corrosion can lead to permanent magnet (PM) demagnetization. When magnets weaken, the generator's voltage drops, and standard controllers attempt to compensate by increasing stator current. This often leads to hitting current limits, causing generator stalls or hardware damage. This paper proposes a robust fault-tolerant control strategy to maintain operational stability during such demagnetization events.
Methodology
The proposed control framework employs a three-layered defense strategy to ensure reliability and performance.
Online Flux Observation (EKF)
To detect demagnetization in real-time without internal sensors, an Extended Kalman Filter (EKF) observer is designed. This algorithm utilizes measurable electrical data—specifically voltage and current—to estimate the actual magnetic flux linkages. By identifying the degree of demagnetization, the system can adapt its control parameters dynamically.
Two-Vector Model Predictive Current Control (MPCC)
Traditional MPCC selects a single optimal voltage vector per control cycle, which can lead to significant current ripple and noise. The authors introduce a two-vector approach that selects two vectors and calculates their optimal acting times (t_opt_1 and t_opt_2). This provides finer control, keeping the stator current within safe limits while accurately following speed commands even under fault conditions.
Improved Luenberger Torque Observer
To distinguish between magnetic faults and external disturbances like wind gusts, an improved Luenberger mechanical torque observer is integrated into the speed control loop. This addition enhances the system's anti-interference capability and reduces speed vibrations caused by unstable wind speeds.
Experimental Results
The strategy was validated using an RT-LAB semi-physical control platform (OP5600) with a 4.5 kW D-PMSG. Key findings include:
- Speed Stability: Unlike traditional PI controllers that fail when current limits are reached during a fault, the proposed method maintained target speed with only ~4% overshoot.
- Dynamic Response: The system recovered from torque changes in 12ms, outperforming the 16ms recovery time of PI control.
- Signal Quality: Total Harmonic Distortion (THD) was reduced from 12.26% (PI control) to 1.18% (proposed strategy).
- Robustness: The system remained stable even with a 50% reduction in magnetic flux (from 0.12 Wb to 0.06 Wb).
Conclusion
The proposed fault-tolerant control strategy effectively mitigates the impact of PM demagnetization in D-PMSG wind power systems. By integrating real-time flux estimation with advanced predictive control, the system achieves superior speed tracking and signal purity. Future work will focus on scaling these results to multi-megawatt turbines and incorporating thermal-based predictive health monitoring.
Summary updated