Sensorless speed estimation of wind driven squirrel cage induction machine

Abstract

Wind Energy Conversion Systems (WECS), among various renewable energy resources, have gained a big boost lately because of being emission free, and relatively more efficient. Nano scale wind turbines operating in category 4 and 5 (wind velocity of 11 to 15 mph) wind speeds are getting a lot of acceptance in coastal areas and remote areas of many countries where such wind speeds are available for most part of a year. For grid integrated systems cage or wound rotor induction machines are commonly used. These are usually controlled by back to back PWM VSI utilizing vector Field Oriented Control (FO). Field orientation requires an accurate knowledge of the rotor speed, which conventionally is measured using tacho-generators, high resolution shaft encoders, and resolvers etc. However, due to relatively high cost of such sensors in low power applications and associated reliability/maintenance issues, sensor-less speed estimation techniques are preferable. Several model based approaches such as MRAS, extended Kalman filtering, Luenberger observers and neural networks have been reported in the literature. These methods provide good speed estimates but have associated high computational cost. Moreover, accuracy of the estimates is also dependent upon the accuracy of the parameters used for motor modeling. With variable environmental conditions, it is difficult to obtain accurate estimates. The method of Current Signature Analysis offers the benefit that the current signature is independent of the machine's time varying parameters. The magnetic saliency related harmonics in the stator current spectrum directly give a measure of the rotor speed. Among these methods, the Rotor Slot Harmonic (RSH) based speed estimation has been reported. Identification of the RSH requires quite complex procedure as its amplitude is approx 2% of the fundamental supply frequency component and it is located in close vicinity of the fundamental. Notch filter based suppression of the fundamental component is a common procedure; but requires the use of high order notch filter which also results in filtering out of useful information in the near vicinity of the fundamental. For an induction machine installed upon a wind turbine, the generated power has additional harmonics due to the dynamics of the wind turbine system e.g. wind shear, tower shadow, blade sampling and the like. Tower shadow especially having significant amplitude (up to 20% of the fundamental in extreme cases) has a detrimental effect on the power quality as it results in Flicker in the generated voltage. This thesis focuses on the harmonic identification and utilization of electrical signature of the wind turbine dynamics specifically the tower shadow effect for sensor-less rotor speed estimation of the generating machine. The machine current signal is sampled and analyzed using well established signal processing techniques to identify the tower shadow harmonic component. Specifically, Hilbert transform is used for amplitude demodulation of the tower shadow signal from the fundamental carrier. FFT based spectral decomposition then gives the required harmonic. This information may be utilized in a number of useful ways for achieving goals like field-oriented control of the machine (PM, reluctance, Induction, DFIG) and/or diagnostics, power quality. By its nature, the method devised and implemented by the author has inherent delays of a few cycles. This limitation, however, can be overcome for applications requiring fast updates (like IFOC) by using model-based observers for speed estimation in conjunction with the proposed scheme for periodic correction which result in zero drift against parameter variations. Moreover, the Flicker effect has been known for power quality degradation at the Point of Common Coupling (PCC) in grid integrated WECS. Different methods have been proposed to mitigate the flicker effect of the wind turbines which include the use of Static VAR Compensators (SVC), Static Synchronous Compensator (STATCOM) and Power System Stabilizers (PSS). This thesis also proposes a method using the tower shadow signature extracted via stator current sampling to mitigate the flicker in the generated output voltage by adding an additional torque loop to the IFOC. The technique is simulated and the results are shown.