Advancements in Control, Design, and Parameter Identification of Permanent Magnet Synchronous Machines

In this work, advancements in control, design, and parameter identification of permanent magnet synchronous machines (PMSMs) are presented. These advancements take up opportunities and challenges offered by novel applications and technologies. In fact, PMSMs are increasingly adopted thanks to high efficiency and versatility, covering a wide range of applications. In particular, the use of PMSMs in the industry of small-scale wind turbines is becoming dominant. In this application, direct drive topologies coupled with variable speed and fixed pitch wind turbines are a promising solution because of their simplicity, efficiency, and reliability. In fact, this solution avoids the use of gearboxes and mechanical systems for the control of the pitch angle, reducing losses, costs, and failure risks. In this architecture, the aerodynamic power regulation is entirely entrusted to the control of the permanent magnet synchronous generator (PMSG). While several methods have been developed to achieve the maximum power point tracking, poor effort has been made to regulate the power for wind speeds above the rated one. In this work, this challenge has been addressed by proposing an innovative control scheme in which an aerodynamic torque observer and a wind speed estimator are involved. The aim of the designed solution is to achieve the maximum power point tracking for wind speeds below the rated one and to extend the power generation in the high wind speed region. In this region, the respect of the safety mechanical limits of the system is the crucial issue. The proposed control scheme has been tested on an experimental setup in the laboratory of Electrical Machines and Drives of the Politecnico di Bari. The achieved results show that the reference power regulation characteristic has been tracked with a good accuracy covering the whole wind speed range of the system. In the context of the design of PMSMs, novel modular topologies are gaining increasing interest, especially in applications where large machines are employed. Thanks to this technology, several advantages in the manufacturing, transporting, and assembling over conventional PMSMs can be achieved. A drawback of PMSMs with modular stators is the presence of additional harmonic components of the cogging torque, with higher amplitudes and lower frequencies than the ones of PMSMs with a one-piece stator. The minimization of these harmonics is essential to increase the control accuracy and reduce undesired noise and vibrations. Despite several methods have been developed for the cogging torque minimization, these mostly deal with conventional PMSMs. In this work, two novel methods to minimize the cogging torque of modular PMSMs are proposed and compared. Both analytical studies and heuristic procedures are adopted to solve the problem. Simulation results with the finite element method have been presented to validate the proposed methods. The results achieved exhibit a reduction of the cogging torque over 90% and show that conventional methods used for one-piece stator PMSMs are not effective against the additional harmonic components produced by modular stators. Also, the spread of the IoTs (Internet-of-Things), edge and cloud computing technologies offers novel perspectives on the monitoring and maintenance of PMSMs. Although data-driven approaches can be considered dominant in this context thanks to the availability of big data, the potential of model-based approaches has not been considered and exploited in this novel scenario. Modelbased approaches are based on the parameter identification of the system. Many well-known solutions have been developed to identify the parameters of PMSMs, but these are not feasible in large-scale applications because these are not designed for straight-through processing where the human component is negligible and IoTs and edge-cloud computing can give their best. Therefore, in the present work, the problem of the automated parameter identification processing data produced by the PMSM drive without ad-hoc tests or control actions is considered. This problem has been addressed by means of the design of an innovative algorithm based on coupled Adaline Neural Networks. Data produced by the PMSM during its ordinary operations are used to feed the Adaline Neural Networks. Moreover, an analytical study has been performed for the convergence and estimation errors analysis of the proposed algorithm. Finally, simulation and experimental investigations have been performed for verification purposes. The results achieved show a good accuracy of the parameter identification, with experimental estimation errors lower than 15% without any manual action.