Real-time Ab Initio Modeling of Electric Machines

电机实时从头建模

• 批准号: 1509804
• 负责人: Ali Davoudi
• 金额: $ 28.5万
• 依托单位: University of Texas at Arlington
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-08-01 至 2021-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1509804&HistoricalAwards=false
• 关键词: Real; time; Ab; Initio; Modeling

Nearly all electricity is produced through electric generators, and 65% of industrial electricity is consumed by electric motors. By 2030, the annual energy consumption of electric motors is expected to exceed 13 quadrillion watt-hours. The ability to undertake optimal design and accurate analysis of electric machines is important to the success of any sustainable energy policy. Weight-critical applications, such as electrified transportation fleets, require radical structural designs with a high power density, high efficiency, and wide operating ranges. To facilitate such aggressive industry-wide transformation, physics-based machine models and high-performance computer simulation tools are indispensable. The fundamental compromise between modeling fidelity and simulation speed for design and analysis of electric machines is investigated in this project. The project is expected to enable a quantum leap in desktop-scale analysis that might not otherwise be achievable until several decades of computational advances lead to the computational capacity needed for real-time analysis of electric machines. First-principle models will be constructed systematically to mimic the characteristics of hardware prototypes, and free the designer from needing expert knowledge and approximations. This is a superior alternative to the existing approaches, namely behavioral approaches that use many simplifying assumptions that compromise model performance, and physical models, which are prohibitively slow. Hybrid order-reduction algorithms and massively parallel hardware-centric tools will accelerate the simulation by up to a million times faster, providing a suitable environment for performance analysis, design optimization, and fault prediction and diagnostics of electric machines. Planned knowledge dissemination of the research results will contribute to multi-disciplinary curricula that help prepare the next generation of power engineers, and to materials that help update the existing workforce to be able to address challenges of the emerging energy conversion devices and systems. The team will engage in STEM-related outreach activities to high school students as well as under-represented groups at UT Arlington, which is a Hispanic-serving institution. The project will investigate an analytical paradigm for electric machines using first-principle physics, and achieve real-time analysis of the resulting ab initio models. The vision is to shift the modeling paradigm away from outdated rule-of-thumb approaches and move toward physics-based yet intuitive models that account for emerging construction materials, computational capabilities, and power electronics-enabled control. The research objectives are to i) construct universal and physical models of electric machines without making heuristic or empirical assumptions common in existing behavioral models, ii) investigate model order reduction algorithms tailored to the resulting ab initio models, iii) map the reduced-order models to massively-parallel hardware-centric simulation platforms to speed up the model execution by as much as six orders of magnitude, and iv) experimentally validate all research findings using prototypes of various machine types commonly used in electric vehicles, electric aircraft, and wind energy conversion systems. This project highlights the fundamental cross-cutting challenges across the domains of power electronics, control systems theory, and scientific computing relevant to electromechanical energy conversion systems. Investigation of such models and tools will enable significantly reduced design cycle time for more efficient electric machines, realistic representation of machine-drive systems for smarter control of energy flow, and co-simulation of temporally-diverse dynamic systems involving both the field equations and grid dynamics for electric machine-to-power grid integration (e.g., in wind farms).

几乎所有的电力都是通过发电机产生的，65%的工业用电是由电动机消耗的。到2030年，电动机的年能耗预计将超过13千万亿瓦时。对电机进行优化设计和准确分析的能力对任何可持续能源政策的成功都很重要。重量关键型应用，如电气化运输车队，需要具有高功率密度、高效率和宽工作范围的激进结构设计。为了促进这种积极的全行业转型，基于物理的机器模型和高性能计算机仿真工具是必不可少的。本项目研究了电机设计与分析中建模保真度与仿真速度之间的基本妥协。该项目有望实现桌面级分析的量子飞跃，除非几十年的计算进步导致电力机器实时分析所需的计算能力，否则可能无法实现。将系统地构建第一原理模型来模拟硬件原型的特征，并将设计师从需要专家知识和近似中解放出来。这是现有方法的一个更好的选择，即行为方法，它使用许多简化的假设，损害模型性能，和物理模型，这是令人望而却步的缓慢。混合降阶算法和以硬件为中心的大规模并行工具将使仿真速度提高100万倍，为电机的性能分析、设计优化、故障预测和诊断提供合适的环境。有计划的知识传播研究成果将有助于多学科课程，帮助准备下一代电力工程师，并有助于更新现有的劳动力，以应对新兴的能源转换设备和系统的挑战。该团队将在德克萨斯大学阿灵顿分校（UT Arlington）向高中生以及代表性不足的群体开展与stem相关的外展活动，这是一所为西班牙裔服务的机构。该项目将利用第一原理物理学研究电机的分析范式，并实现对从头算模型的实时分析。其愿景是将建模范式从过时的经验法则方法转变为基于物理的直观模型，这些模型考虑了新兴的建筑材料、计算能力和电力电子支持的控制。研究目标是：i)构建电机的通用和物理模型，而不做现有行为模型中常见的启发式或经验假设；ii)研究针对生成的从头算模型定制的模型降阶算法；iii)将降阶模型映射到大规模并行的以硬件为中心的仿真平台，以加快模型执行速度多达6个数量级。iv)使用电动汽车、电动飞机和风能转换系统中常用的各种机器类型的原型，对所有研究结果进行实验验证。该项目强调了与机电能量转换系统相关的电力电子、控制系统理论和科学计算领域的基本交叉挑战。对这些模型和工具的研究将大大缩短更高效电机的设计周期时间，为更智能的能量流控制提供机器驱动系统的真实表示，以及对涉及电场方程和电网动力学的时间变化动态系统的联合模拟，以实现电机与电网的集成（例如，在风力发电场）。