Predictive control approaches to fault tolerant control of wind turbines
Thesis or dissertation
- © 2014 Xiaoran Feng. All rights reserved. No part of this publication may be reproduced without the written permission of the copyright holder.
This thesis focuses on active fault tolerant control (AFTC) of wind turbine systems. Faults in wind turbine systems can be in the form of sensor faults, actuator faults, or component faults. These faults can occur in different locations, such as the wind speed sensor, the generator system, drive train system or pitch system. In this thesis, some AFTC schemes are proposed for wind turbine faults in the above locations. Model predictive control (MPC) is used in these schemes to design the wind turbine controller such that system constraints and dual control goals of the wind turbine are considered. In order to deal with the nonlinearity in the turbine model, MPC is combined with Takagi-Sugeno (T-S) fuzzy modelling. Different fault diagnosis methods are also proposed in different AFTC schemes to isolate or estimate wind turbine faults.
The main contributions of the thesis are summarized as follows:
A new effective wind speed (EWS) estimation method via least-squares support vector machines (LSSVM) is proposed. Measurements from the wind turbine rotor speed sensor and the generator speed sensor are utilized by LSSVM to estimate the EWS. Following the EWS estimation, a wind speed sensor fault isolation scheme via LSSVM is proposed.
A robust predictive controller is designed to consider the EWS estimation error. This predictive controller serves as the baseline controller for the wind turbine system operating in the region below rated wind speed.
T-S fuzzy MPC combining MPC and T-S fuzzy modelling is proposed to design the wind turbine controller. MPC can deal with wind turbine system constraints externally. On the other hand, T-S fuzzy modelling can approximate the nonlinear wind turbine system with a linear time varying (LTV) model such that controller design can be based on this LTV model. Therefore, the advantages of MPC and T-S fuzzy modelling are both preserved in the proposed T-S fuzzy MPC.
A T-S fuzzy observer, based on online eigenvalue assignment, is proposed as the sensor fault isolation scheme for the wind turbine system. In this approach, the fuzzy observer is proposed to deal with the nonlinearity in the wind turbine system and estimate system states. Furthermore, the residual signal generated from this fuzzy observer is used to isolate the faulty sensor.
A sensor fault diagnosis strategy utilizing both analytical and hardware redundancies is proposed for wind turbine systems. This approach is proposed due to the fact that in the real application scenario, both analytical and hardware redundancies of wind turbines are available for designing AFTC systems.
An actuator fault estimation method based on moving horizon estimation (MHE) is proposed for wind turbine systems. The estimated fault by MHE is then compensated by a T-S fuzzy predictive controller. The fault estimation unit and the T-S fuzzy predictive controller are combined to form an AFTC scheme for wind turbine actuator faults.
- Department of Engineering, The University of Hull
- Patton, Ron, 1949-; Hou, Ming
- Sponsor (Organisation)
- University of Hull; China Scholarship Council
- Qualification level
- Qualification name
- 3 MB