电力电子装置在能源领域正得到广泛应用，而工业仿真软件则是电力电子装置设计和分析的底层支撑工具。然而，电力电子系统是一个典型的连续与离散混杂系统。大容量装置中开关数目多、开关频率高，且运行场景复杂多样。这些特征使得现有的仿真工具在求解电力电子系统时存在严重局限性，尤其是仿真大规模电力电子系统和小时间尺度开关瞬态过程时，动辄耗时数小时甚至几天，大大限制了软件的实际应用。离散状态事件驱动（Discrete State Event-Driven，DSED）仿真框架是近年来提出的一种针对电力电子系统的高效求解方法。该方法以事件驱动、状态离散代替了传统的时间离散驱动机制，契合电力电子混杂系统的本质，展示出极大的应用潜力。本文基于DSED的已有成果展开工作，更加关注工业仿真软件“工业”和“软件”的双重属性，从通用仿真软件计算机自动化实现流程的角度出发实施电力电子系统建模解算方法的理论创新，进而研发通用的电力电子工业仿真软件。首先，面向电力电子系统连续与离散混杂的特点，针对开关动作导致拓扑结构频繁变化、方程更新的时空复杂度高的问题，提出了基于基本桥臂的开关函数受控源等效建模方法以及半符号化的状态方程更新方法，解决了电力电子系统方程生成及更新耗时极长的瓶颈问题，为将DSED框架发展为通用仿真软件奠定了基础。其次，针对大规模电力电子系统求解耗时长的问题，提出了电路划分的计算机自动实现方法。基于电路划分结果，推导证明了电力电子系统状态方程的分块稀疏性质。应用该性质，提出了能够适用于非刚性/刚性系统的稀疏优化求解方法，从而提升了工业仿真软件自动化求解各类大规模电力电子系统的仿真速度。然后，针对仿真复杂场景的实际应用需求，提出了基于刚性检测的电力电子系统自适应仿真算法。该方法可以在全时间域上始终选择最优求解器进行计算，从而提升了软件求解各类复杂电力电子系统时的综合仿真性能，增强了仿真设置的自适应性和用户友好性。最后，所提方法结合DSED已有成果，在首款国产电力电子工业仿真软件DSIM中得以实现。在多个实际装置算例研究中，DSIM相较于国际先进软件仿真速度平均提升两个数量级以上，其中本文所提方法综合贡献数十倍提速，仿真结果与实验波形相吻合。DSIM软件已经作为高效数值实验平台支撑装置的设计与分析。
Power electronic converters are widely used in the energy field, and industrial simulation software is the underlying support tool for their design and analysis. However, power electronic systems are typical hybrid systems consisting of both continuous dynamics and discrete events. In high-power systems with numerous switching devices and high switching frequency, the operating scenarios are very complicated. These features impose serious limitations on existing simulation tools, especially when simulating large-scale power electronic systems and switching transients with small time-scale. The simulation process takes hours or even days, which greatly restricts the practical applications of the software.The discrete state event-driven (DSED) simulation framework is an efficient tool proposed in recent years for power electronic systems. This method adopts event-driven and state-discretized algorithm instead of the traditional time-driven and time-discretized one, which fits the nature of hybrid power electronic systems and demonstrates great application potentials. This paper inherits the existing DSED framework, but focuses more on both the “industrial” and “software” attribute of industrial simulation software. With consideration of the computer automation process in general-purpose simulation software, novel modeling and solving methods for power electronic systems are proposed and industrial simulation software is developed.Firstly, considering the hybrid nature of power electronic systems, the switching events will cause the frequent changes of circuit topologies so that the time-and-space complexity of equation updating can be very high. A switching-function controlled-source model for basic switching legs and an automated semi-symbolic state equation updating method are proposed to solve the bottleneck of the time-consuming equation generation and updating process, which makes it possible to develop the DSED framework into general simulation software.Secondly, to deal with the problem of the long solving time for large-scale power electronic systems, an automatic circuit partitioning strategy is proposed. Based on the partitioning results, the block sparsity of the state equation matrices for power electronic systems can be derived and proven. Furthermore, the property is applied to the sparsity optimization for both non-stiff and stiff systems. Thus the speed of industrial software when simulating various large-scale power electronic systems can be improved.Thirdly, an adaptive simulation method for power electronic systems based on stiffness detection is proposed to meet the requirements of simulating complicated scenarios. It can always select the better numerical solver during the whole simulation process, so that the overall computational efficiency is improved. The self-adaptability of simulation settings and the user-friendliness are enhanced as well.Finally, combined with the existing DSED approach, the proposed methods have been implemented in the first domestic power electronic industrial simulation software DSIM. In multiple case studies of practical power converters, DSIM can achieve about hundred-fold acceleration on average compared with international advanced software, where the methods proposed in this paper contribute several-dozen-fold speedup independently. The simulation results also match well with the experiments. In total, DSIM has worked as an efficient numerical experimental platform which can support the design and analysis of real power electronic systems.
