One problem arising from re-defining the fault signal is that it Anguizole is not easy to decide the severity of a fault as it is strongly coupled to the system states. The real fault signal can be constructed using the approach proposed in Section 2.4. The reconstructed fault is shown in the Fig 13 from which it is very easy to determine whether there is a fault or not and also the severity of the fault. The result given in Fig 13 corresponds to the case of oil with abnormally high air content – with expected fault severity of θf = 1.
Fig. 13. Fault θf and its estimate in the high air content case.Figure optionsDownload full-size imageDownload as PowerPoint slide
4.2. Actuator fault in generator subsystem
The estimation of the torque actuator fault is presented in Fig 14(a). It can be seen that the LPV ESO method can provide very good fault estimation, which is a significant result even though it is claimed that a 100 Nm fault is too small to be detected . From the simulation results, there is no obvious improvement obtained by using the AFTC scheme. However, generator torque offsets will result in large amount of lost energy production over time. In addition, tt is important to point out that the actuator fault should be detected as early as possible to prevent the impact of faults from other subsystems, or even to prevent a gross effect on the overall system performance. To test the robustness of the fault estimation at different operation points, another scenario that the fault occurred between 3100s and 3200s is carried out and the simulation result is given in Fig 14(b).