Nonlinear Control and AI-Driven Strategies for Enhancing Performance in Permanent Magnet Synchronous Motor Drives

Abstract

이 논문에서는 영구자석 동기기(permanent magnet synchronous machine, PMSM)의 제어 특성을 향상시키기 위하여, 비선형 제어 및 인공지능(artificial intelligence, AI)기법을활용한제어전략을제시한다.현재까지영구자석동기 모터의운전을위해서는대부분비례-적분(PI)제어기를활용한선형직렬제어 방식을 사용해왔다. PI 제어는 단순하고 구현하기 쉽기 때문에 여전히 산업용 애플리케이션에서사용되고있지만구동인버터의비선형성으로인해외란또 는파라미터변화에따라 PI제어기의성능이쉽게저하되는문제가있다.또한 PI 제어기는 포화특성이 없는 선형 시스템에 적합하다. 제어기의 성능은 플랜 트모델에의존적이며좁은범위내에서우수한성능을위해조정되므로넓은 운전영역에서 훌륭한 폐루프 성능을 유지하는 것은 어려운 과제이다. 이러한 문제들을극복하기위해이연구에서는 meta-heuristic approach를구현하여정 확한플랜트모델없이도제어파라미터를조정할수있는방법을제시한다. 지난몇년동안 PMSM제어전략에괄목할만한발전이있었다.비선형제 어설계를통하여폐루프제어성능을향상시키고과도상태및정상상태동작 조건에서 강인성을 높이려는 노력들이 이어져왔다. 또한 최근에는 지능형 제 어설계개발을위해 AI기반제어전략들이빠른속도로소개되고있다.이러한 제어설계 기법들은 시스템의 비선형성 및 제약 조건을 하에서 시스템을 효과 적으로구동할수있게해준다.뿐만아니라지능형제어설계는플랜트모델에 대한 사전 지식이 필요하지 않으며 네트워크의 학습이 오프라인에서 이루어 지고학습된네트워크가온라인으로구현되기때문에일부비선형제어전략에 비해실시간계산에대한부담이덜하다. 이 논문은 두 부분으로 구성된다. 첫 번째는 제어전략 설계의 주요 동기와 문헌에서구현된제어전략들을소개하고 PMSM드라이브의수학적모델링에 대해자세히설명한다.두번째는 PMSM드라이브의우수한동적성능을위한 고급 제어전략의 설계 및 적용 방법을 제시하며, 총 3개의 장으로 구성된다. 첫번째장에서는플랜트모델없이선형제어기이득을미세조정하는최적화 알고리즘을제시한다.이 model independent tuning approach는기존선형제어 튜닝방식보다우월한동적성능을제공한다.두번째장에서는 PMSM드라이 브를위한고급비선형제어기법을제안하며,제안된제어전략은 reaching law approach기반의 sliding mode control을이용한다.세번째장에서는 PMSM구 동시스템에대한인공지능(AI)의적용에대해다룬다.선형또는비선형제어 설계를대체하고사전전문지식없이도우수한성능을제공할수있는몇가지 최첨단제어설계기법을제안한다.마지막으로실험및시뮬레이션을통하여 제안된 기법들의 적합성을 평가하며, 제시된 테스트 결과들을 통하여 제안된 제어 설계 기법이 PMSM 구동에 있어서 향상된 동적 특성을 보여줌을 증명 한다.|This thesis presents adequate control strategies for PMSM drive control. Different control schemes have been proposed over the years to enhanced control performance of motor drive system. In past the most of the control strategies are based on traditional linear control scheme that is proportional-integral(PI) control for cascaded control design of motor drive. PI controls are still employed in industrial application due to there simplicity and easy to implement. But due to nonlinear nature of power electronic systems with multiple inputs, and outputs the performance of PI can easily be degraded under disturbance or parametric variations. Moreover, PI controllers are suitable for linear systems with unconstrained control problems. The controller performance is highly dependent on plant model, where they are tuned for excellent performance within narrow working operating range, thus possessing excellent closed-loop performance is still challenging. To overcome this challenge the meta-heuristic approaches are implemented in this work to tune control parameters without need of the plant model. However, over the wide operating range the dynamic performance is still poses practical challenges.

In recent years, there is remarkable advancement in control strategies for Permanent Magnet Synchronous Machine(PMSM). Nonlinear control designs are implemented to enhanced the closed-loop control performance and provide robustness in transient and steady-state working operating conditions. Also, nowadays AI-based control strategies are emerging at fast-pace for the development of intelligent control design. Moreover, these control design are capable of handling system nonlinearities and constraints with effective manner. Furthermore, these intelligent control design doesn't require prior knowledge of plant model as well as computationally less expensive then some nonlinear control strategies because the training of the network is done offline and the trained network is implemented online.

This thesis is divided into two parts. The key motivation behind the design of control strategies and the implemented control strategies in the literature are presented in the first part. Furthermore, the modeling of mathematical model of the motor drive implemented in this thesis is presented in detail. The second part presents the design and application of advanced control strategies for excellent dynamic performance of motor drives, and is divided into three chapters. The first chapter demonstrate the optimization algorithm to fine tune the linear controller gains for efficient dynamic performance without the need of plant model. The model independent tuning approach give optimized linear control superiority over traditional linear control. The second chapter is devoted to the design of advanced nonlinear control for the application motor drives. The chapter will demonstrate the reaching law sliding mode control for enhanced dynamic performance at low speed operation with effective compensation technique. This provided approach ensure excellent speed tracking performance and remain robust in presence of uncertainties and mitigate the undesirable high-frequency chattering resulting from the rapid mode switching within the control system, which aims to maintain the state on a predefined sliding surface and consequently causes abrupt and oscillatory fluctuations in the control input. The third chapter address the application of Artificial intelligence(AI) for motor drive system. Some state of the art control design are proposed that can replace Linear or non-linear control design and give excellent performance without prior expert knowledge. Finally, this thesis thoroughly evaluates motor drive performance using a combination of experimental and simulation-based tests. The presented test results help improve motor drive systems by demonstrating exceptional performance features.