大容量电力电子系统已经广泛应用于现代电网等重要工业领域。其中，建模仿真是进行分析设计和控制运行不可或缺的基础方法和工具。然而，大容量电力电子系统是典型的大规模“连续-离散”混杂系统，时间尺度跨越秒到纳秒等多个数量级，其动力学表征目前面临严重的理论瓶颈，表现为：微纳秒级开关过程建模解算发散，无法支撑系统小时间尺度瞬态分析和安全可靠运行；仿真大规模系统经常耗时几小时甚至几天，以致难以对复杂系统进行优化设计和精准分析控制。为此，本文从多时间尺度混杂系统的角度出发，针对复杂电力电子系统建模仿真理论方法开展研究，提出了一套解耦型离散状态事件驱动建模仿真方法，形成了大容量电力电子混杂系统的多时间尺度动力学表征分析方法。首先，提出了广义混杂系统的概念，将传统混杂系统提升为离散事件包含瞬变过程的多时间尺度系统；基于此提出了解耦型离散状态事件驱动理论方法，在建模上实现了多时间尺度解耦表征，在解算上实现了以状态离散代替时间离散、以事件驱动仿真进行的高效机制，为广义混杂系统的仿真认知奠定了理论基础。其次，面向小时间尺度开关过程建模，提出了分段解析瞬态模型，实现了基于时间分段、机理解耦和参数解耦的瞬态动力学表征；以此为基础，面向多时间尺度“连续-离散”过程建模，提出了分层自动机模型，在解耦型离散状态事件驱动理论框架下形成了数值稳定的多时间尺度动力学表征。然后，面向大规模系统连续状态仿真解算，提出了解耦型状态离散算法，通过子系统解耦解算和基于状态离散的接口误差控制，实现了相同精度下解算速度的大幅提升；进一步，面向离散事件及其瞬变过程仿真解算，提出了基于事件驱动的多时间尺度解耦算法，解决了发散问题，在解耦型离散状态事件驱动理论框架下形成了准确、快速、稳定的仿真解算方法。以此为基础，进行了多个不同规模和类型的仿真算例研究，实现了相同精度下平均两个数量级（100倍）以上的仿真速度提升，可建模求解的时间尺度拓展到微纳秒级，仿真结果与实验结果吻合。最后，面向开关过程瞬变电磁场建模仿真，提出了多时间尺度-多物理域解耦方法，在超算平台上解算了纳秒级关断过程空间电磁场的变化，通过电磁实验测量验证了仿真结果的有效性，在此基础上可视分析了关断过程电磁能流的瞬变现象和规律，为研究开关过程电磁瞬变机理提供了基础方法。
High-power electronics systems have been largely integrated in many key industrial areas such as modern power grid. Modeling and simulation acts as an indispensable tool for the analysis, design, control and operation of such systems. However, in high-power electronics systems, continuous states and discrete events coexist and interact with each other, and the timescale varies from second to nanosecond level. These features all contribute to a serious bottleneck in terms of system modeling and simulation. For one thing, the simulation of micro/nanosecond level switching transient faces the convergence problem, making it difficult or even impossible in analyzing transient behaviors. For another, the simulations of large systems composed of hundreds or thousands of switching devices usually take hours or even days, restricting the optimal design and accurate analysis and control of such complex systems.Therefore, from the perspective of hybrid system, this thesis studies the modeling and simulation of high-power large-scale complex power electronics systems, establishes a decoupled discrete state event-driven approach, and forms the multi-timescale dynamic analysis framework for high-power electronics hybrid systems.Firstly, the concept of generalized hybrid system (GHS) is proposed, extending the conventional concept of hybrid system towards the multi-timescale system considering the transients within discrete events. Then a decoupled discrete state event-driven (DSED) approach for GHS is proposed. In system modeling, the multiple timesclaes are decoupled; and in system solving, the time-discretized and time-driven algorithm is transformered into an efficient state-discretized and event-driven one, which lays a theoretical foundation in this thesis.Secondly, in terms of small-timescale switching transient modeling, a piecewise analytical transient (PAT) model is proposed, which describes the micro/nanosecond-level switching transient based on time slicing, mechanism decoupling and parameter decoupling. Then in terms of the modeling of multiple timescales, a hierarchical automaton for the multiple timescales is proposed which constitutes a numerically stable decoupled multi-timescale model under the decoupled DSED framework.Thirdly, in terms of large-scale system solving, a decoupled discrete state algorithm for the integration of continuous states is proposed, which achieves a decoupled solving of the large-scale system and the effective error control of the interface, in together largely increasing the simulation speed with complete accuracy. Then in terms of the solving of the discrete event together with its switching transient, a timescale decoupling algorithm is proposed which eliminates the convergence problem. In together, an accurate, fast and stable numerical solver is established under the decoupled DSED framework.The proposed modeling and simulation approaches are verified in multiple case studies of different scales and categories. In general, a two-order-of-magnitude (100 times) acceleration is achieved, with no convergence problem, and the solvable timescale is extended towards micro/nanosecond level. The simulated results coincide with the experimental results.Finally, the proposed decoupled approach is extended towards the simulation of electromagnetic filed. A multi-timescale and multi-physical-domain decoupled model is proposed, and the transient electromagnetic filed during a switching-off process is successfully calculated on a supercomputer. Electromagnetic measurement experiments are performed to verify the correctness of the simulation, and based on the simulated results, the transient distribution and variation of the electromagnetic energy flow during the switching-off transient is visualized and analyzed, providing a basic tool to study the electromagnetic mechanisms during switching transient.
