Collaborative Research: Digital Twin Predictive Reliability Modeling of Solid-State Transformers

固态Transformer（SST）被认为是未来电力系统的革命性技术。它比传统的电磁式Transformer结构紧凑，在潮流控制和电能质量调节方面具有显著的可控性优势。然而，限制SST实用性的一个主要技术障碍是与传统变压器相比可靠性低。这是由于包括半导体晶体管、辅助电路和无源元件在内的大量器件。目前，SST的可靠性很少受到关注，这限制了其商业化和工业采用。该项目将为SST开发数据驱动的数字孪生模型，这将有助于预测组件退化和预防灾难性故障。这旨在显著提高SST在安全关键应用中的可靠性，例如未来的电力系统和电气化运输应用。提出的建模和设计方法将导致新的电力电子设计工具，并将使一个完全集成的设计过程，将产生新的拓扑结构，节省大量的设计和实施时间。此外，这些方法将增强可靠性建模，即使是新合成的架构，也可以在设计阶段准确地估计可靠性。关于教育影响，这项工作提供了一个机会，将人工智能应用于电力电子工程。因此，本项目的成果将提升电力电子教学课程，并为学生提供未来电力工程的有效技能。为了应对SST的可靠性挑战，本项目将开发一个全面的SST在线健康监测系统框架，以显着提高面对电气故障的可靠性。拟议的健康监测框架将包括在线预测和诊断SST可能遇到的潜在电气故障，针对常见的半导体开关故障和高频变压器的健康退化。具体而言，将通过与电力电子构建模块集成的智能栅极驱动器来监控关键SST参数的组合，因此可以在故障初始阶段预测和诊断半导体模块的退化。提出了一种新的数据驱动数字孪生方法来预测SST转换器模块的行为，并且它将计算健康性能指标，以使该技术与完整物理模型计算相比在计算上更高效。快速在线诊断算法将被开发并嵌入SST微控制器中，因此可以识别和描述故障，以最大限度地减少停机成本并避免级联故障。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Solid state transformer (SST) is deemed as a revolutionary technology for future power systems. It is much more compact than the conventional electromagnetic transformer, with significant advantages of controllability both in power flow control and power quality regulation. However, one major technical barrier that constrains the practicality of SST is the low reliability compared to the conventional transformers. This is due to the large device count including semiconductor transistors, auxiliary circuits, and passive components. Currently, the reliability of SST has received little attention, which constrains their commercialization and adoption by industry. This project will develop data-driven digital twin models for SSTs that will facilitate prediction of component degradation and prevention of catastrophic failures. This is aimed to significantly improve the reliability of SSTs for safety-critical applications, such as future power systems and electrified transportation applications. The proposed modeling and design methods will result in new classes of power electronics design tools and will enable a fully integrated design process that will generate new topologies and save substantial design and implementation time. Further, these approaches will enhance reliability modeling where reliability can be accurately estimated from at design stage even for newly synthesized architectures. Regarding educational impact, this work presents an opportunity to apply artificial intelligence to power electronics engineering. Hence, the outcome of the project will upgrade power electronics teaching curricula and provide students with an effective skillset for future power engineering.To address the challenge of reliability of SSTs, this project will develop a comprehensive systematic framework of online health monitoring for SSTs to significantly improve the reliability in the face of electric faults. The proposed health monitoring framework will include online prognosis and diagnosis of potential electrical faults that SSTs could be subject to, targeting common semiconductor switching faults and health degradation in high-frequency transformers. Specifically, a portfolio of critical SST parameters will be monitored through a smart gate driver that will be integrated with the power electronic building blocks, so degradation in the semiconductor modules can be predicted and diagnosed during the fault inception stage. A novel data-driven digital twin approach is proposed to predict the behavior of the SST converter modules and it will compute health performance indices to make the technique more computationally efficient compared to full physical model computations. Fast online diagnostic algorithm will be developed and embedded in the SST microcontroller, so a fault can be identified and characterized, to minimize downtime cost and avoid cascading failures.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.