Multilevel direct torque control of sensorless two five-phase induction machines connected in parallel

Abstract

Railway traction, naval propulsion, and robotics are just a few examples of numerous industrial sectors that exploit the multi-machine drive, more than one electrical machine drive. Which is currently comprising n three-phase machines parallel-connected in different configurations. However, it lacks the means of independent control of each machine in the drive due to the Master-Slave and Averaging control schemes utilized in the three-phase system. Therefore, this thesis considers the concept of an independently controlled two-machine drive constituted of two multiphase induction machines (IM) parallel-connected to a single multi level multiphase voltage source inverter (VSI), to enhance the concept of multi-phase multi machine drive when compared to its three-phase counterpart. Where the five-phase IM (FPIM) topology is proposed based on a cost-effective technique study, by exploiting the additional degrees of freedom (DOF) offered by this topology. The fundamentals of the independent control concept of the parallel operation of the two-machine drive is derived from the fact that the AC machines, disregarding the phase number, require the control of only one set of current components. Thus, the extra current set can be exploited to control independently the other machines in the drive with the proper stator phase connection. Nevertheless, this concept faces numerous challenges due to the inherited control difficulties of the IM and the multi-level topologies of the VSI, the parallel operation complexity, and the poor performance of the traditional control strategies, especially during the low-speed operation. For this reason, the direct torque control (DTC) technique presents an interesting solution to overcome the aforementioned challenges, offering robust and dynamic performance. Nevertheless, every technique has its pros and cons; for instance, it presents a high content of flux/torque ripples, in addition to the size and complexity of the switching table (ST), which reduces the performance of the suggested concept of the two-machine drive. For this reason, an AI-based DTC technique, exploiting the artificial neural networks (ANN) benefits, to improve xx Abstract the robustness and enhance the performance of the drive resulting in the reduction of the developed flux/torque ripples. This latter is achieved by replacing the hysteresis controllers (HC) and the ST of the conventional method by ANNs-based controller; thus, resulting in a simple yet more robust structure. Moreover, to further optimize the performance, size and improve the aforementioned cost effective technique study based two-machine drive’s fault-tolerant capability, Machine Model based (MM) sensorless methods are investigated to observe and estimate the machine’s critical state variables to achieve the optimum performance of the DTC technique and the drive by addressing the drawback of the utilized open-loop estimator especially during the low-speed operation. Where the three proposed MM techniques based on Model reference adaptive system (MRAS), Sliding mode observer (SMO), and Virtual current sensor (VCS) results in higher performance regarding the speed tracking during the transient and steady state operation, negligible estimation error, and better rejection of external disturbances in different operating scenarios. On the other hand, a preventive mitigation approach to tackle the Common-Mode Voltage (CMV) generated by the PWM-controlled VSI’s discrete voltages is considered. For this reason, a modified DTC is discussed, utilizing the concept of virtual voltage vector (VVV). These VVVs are synthesized based on a selected switching states capable of reducing the CMV peak to-peak values. Thus, improving the drive’s reliability and reducing the maintenance’s time and cost by addressing the inherited drawback of the multi-level VSI from its roots.