Prognostics and health aware model predictive control of wind turbines

Résumé

Wind turbines components are subject to considerable stresses and fatigue due to extreme environmental conditions to which they are exposed, especially those located offshore. Also, the most common faults present in wind turbine components have been investigated for years by the research community and that has led to propose a fault diagnosis and fault tolerant control wind turbine benchmark which include a set of faults that affect the sensors and actuators of several wind turbine components. This thesis presents some contributions to the fields of fault diagnosis, fault-tolerant control, prognostics and its integration with wind turbine control which leads to proposing a control approach called health-aware model predictive control (HAMPC). The contributions are summarized below: - Model-based fault diagnosis: to perform fault detection and isolation interval-based observers together with a set of analytical redundant relations (ARRs) are obtained based on a structural analysis and the fault signature matrix that relates the ARRs with the faults. - Fault tolerant control: it is proposed a fault tolerant control scheme that integrates fault detection and an algorithm for fault accommodation. The scheme has the objective to avoid the increment of blades and tower loads when a fault in the rotor azimuth angle sensor occurs using the individual pitch control technique (IPC). - Wind turbine blades fatigue prognostics and degradation: fatigue is assessed using the rainflow counting algorithm which is used to estimate the accumulated damage and for degradation, it is used a stiffness degradation model of blades material which is used to make predictions of remaining useful life (RUL). - Wind turbines health control: the module for the health of the system based on fatigue damage estimation and RUL predictions is integrated with model predictive control (MPC) leading to the proposed control approach (HAMPC). The contributions presented in this thesis have been validated on a wind turbine study case that uses a 5MW wind turbine reference model implemented in a high fidelity wind turbine simulator (FAST).