Condition Monitoring the Drive Train of a Direct Drive Permanent Magnet Wind Turbine Using Generator Electrical Signals

The benefit of wind turbine (WT) can be significantly improved through a well-organized condition-based maintenance strategy. However, such a target has not been fully achieved today. One of the major reasons is lack of an efficient WT condition monitoring system (CMS). The existing WT CMSs often involve high initial capital cost, with complex structure, suffer from inefficient management and show unsatisfactory hardware reliability. So, the operators still have desire for an economical, effective, and reliable CMS for their machines. The work reported in this paper is intended to meet such a demand. Because direct drive permanent magnet (PM) WTs are showing increased market share, but the existing WT CMSs are not designed to deal specifically with this new design, this paper reports on a CM technique dedicated to monitoring the drive train of direct drive WTs. Instead of taking the vibration analysis approach that is being popularly adopted by commercial WT CMSs, a novel CM strategy is researched in this paper by introducing generator electrical signals into WT CM and interpreting them by using a dedicated criterion named instantaneous variance (IV) and Teager–Huang transform (THT), i.e. the generator electrical signals will be evaluated first by using the IV, of which the fault detection capability can be further enhanced with the aid of empirical mode decomposition (EMD). Once an abnormality is detected, then detailed THT analysis of the signal will be conducted for further investigation. The technique has been verified experimentally on a specifically designed WT drive train test rig, on which a PM generator rotates at slow variable speed and is subjected to varying load like a real WT does. Considering the electric subassemblies and rotor blades of direct drive WTs are most vulnerable to damage in practice, rotor unbalance and generator winding faults were emulated on the test rig. Experimental results show that the proposed CM technique is effective in detecting both types of faults occurring in the drive train of direct drive PM WTs. In summary, the proposed CM technique can be identified by (i) the CM is accomplished through analyzing the generator electrical signals without resorting to any other information (e.g. vibro-acoustic). Hence, the data acquisition work will be eased off; (ii) no more transducer other than current and voltage sensors are required. Thus, the cost of the CMS will be significantly reduced; (iii) attributed to the distinguished superiorities of THT to traditional spectral analyses in processing nonlinear signals, the proposed technique is more reliable in interpreting WT CM signals; and (iv) the CM criterion IV has a simple computational algorithm. It is therefore suited to both online and offline WT CM applications.