Induction machines (IMs) are the driving force in industries such as manufacturing, transportation and wind power generation. Hence, it is essential to reliably detect faults in IMs so as to enhance production quality in manufacturing and avoid operational degradation. However, it is still challenging to reliably detect faults in IMs as fault feature properties could change under variable IM operating conditions. The first objective of this thesis is to develop an enhanced empirical mode decomposition (EEMD) technique to detect an IM broken rotor bar (BRB) fault based on motor current signature analysis. In the developed EEMD technique, a phase insensitive similarity function is initially suggested to determine the representative intrinsic mode functions (IMFs). Moreover, an optimized adaptive multi-band filter is suggested to process the current spectrum and to recognize the fault characteristic features. Likewise, a modified whale optimization algorithm (MWOA) is proposed, which is utilized to optimize the parameters in adaptive multi-band filter. Finally, a reference function is recommended to enhance feature properties and IM fault detection. The effectiveness of the proposed EEMD technique is verified through experimental analysis under different IM operating conditions.

Different arrangements of IMs are predominantly used in wind energy conversion (WEC) systems as IMs can easily operate at variable wind speed and can be connected simply with the grid. However, it is challenging to control a healthy IM based WEC system due to wind speed uncertainties and nonlinearity of the IMs as well as WEC system. Among the different configurations of IMs, the doubly fed induction generator (DFIG)-based wind energy conversion (WEC) system is the mostly used configuration due to its decoupled nature of active and reactive power control. Since the traditional DFIG-based WEC systems are controlled using proportionalintegral (PI) controllers, they are not always capable of coping with wind uncertainties. The second objective of this work is to propose a new adaptive neuro-fuzzy (ANF)-based direct torque and flux (DTF) control technique for grid-connected DFIG WEC systems. This new control technique has better dynamic responses to handle the uncertainties of the wind speed variation. A training algorithm is also established to adaptively optimize the ANF system parameters to accommodate system nonlinearities and wind speed uncertainties. The proposed ANF-based DTF control technique improves the dynamic performance of the WEC system compared to the conventional PI-based DTF control technique, as verified by simulation and experimental results.

Furthermore, the third objective of the thesis is to develop an integrated neuro-fuzzy (INF)- based DTF control technique for DFIG-based WEC system. Traditionally for DTF control technique, two separate neuro-fuzzy (NF) structures are utilized to control torque and flux independently. In this work, a single INF structure is developed to control torque and flux simultaneously, as these are interdependent through rotor side converter of the DFIG. A new training method based on ensemble algorithm is developed for online training of the proposed INF system parameters. The proposed INF-based DTF control technique is implemented in real time for a prototype DFIG-based WEC system in a laboratory environment using DSP board DS1104. The efficacy of the proposed INF-based DTF control technique, incorporated with the ensemble training method, is investigated and compared with the classical NF network and training method. The proposed INF-based DTF control technique improves the dynamic performance of the WEC system compared to the conventional NF-based DTF control technique . Furthermore, the INF network has less computational burden and adaptive capability compared to its counterpart. The stability of the proposed INF-based DTF control technique is also analyzed under variable wind speed conditions. The WEC system with the proposed INF-based DTF control technique is found to be stable under dynamic conditions.

The squirrel cage induction machine (SCIM) is also utilized sometimes as wind generator besides DFIG as well as used largely in industrial applications. However, different from DFIG, the BRB fault may occur in SCIM which may affect the operating performance of the machine. The broken rotor bars can cause current and torque fluctuation of IM which can adversely affect the operational condition of the machine. Hence, the fourth objective of the thesis is to develop a NF based fault tolerant control technique, which can reduce the torque ripple at the time of broken rotor bar fault of IM and improve the operational performance. The effectiveness of the proposed FTC technique is also investigated in this thesis.