Prof. Dr. Drazen Dujic
Head of Power Electronics Laboratory, Ecole Polytechnique Fédérale de Lausanne
CASE STUDY Accelerate power electronics development - Choose your tools wisely
Power electronics development tools have come a long way forward, thanks to affordable and ever-increasing, performance wise, digital signal processing solutions. Modeling and simulations, both off-line and more importantly, in the real-time, are essential steps in the development of power electronics solutions.
Almost ten years ago, I was lucky to be member of the R&D team working on the development of the world’s first 1.2MW Power Electronics Traction Transformer (PETT) demonstrator. The demonstration project was including the 12 months field trial in Swiss railway network. Nowadays, this technology is branded differently as the Solid State Transformer (SST) and it is still found attractive by both academia and industry. Nevertheless, the project was exciting and fast paced, as pilot demonstration of the PETT prototype was awaiting immediately after the development and laboratory testing.
One of the first tasks I was given was to build a replica of the system that could be used to support control software development. Essentially, this meant downscaling complete power hardware of the system, so that existing control boards and new control software under development could be tested with reduced risks. In those days, analogue simulators were still dominant solution for this problem, with multiple of them being available for R&D engineers for the standard products and platforms. Digital simulators or Real-Time Hardware-in-the-Loop (RT-HIL) systems were still perceived novelty in power electronics, despite being used in power system domain for many years. Faced with the technical challenges of task, time pressure of the project schedule, budget constraints, risks associated with new RT-HIL systems, decision was made to build an analogue simulator, the one shown in Fig. 1.
Fast forward few years later, with industry years behind me and now leading university research laboratory, I was faced with the similar problem. For our own research needs, we were developing medium voltage Modular Multilevel Converter (MMC). The needed effort was very high, especially considering that involved personnel were all PhD students. Young, motivated, and hard-working, but not necessarily experienced or ready for the complexity of tasks ahead of them. To reduce the risks during developments, and make it possible to work in parallel on the hardware and software of the converter, a replica of the future hardware was needed. This time, however, decision was made to pursue the path of RT-HIL adoption as support for control software part of the project, as illustrated in Fig. 2. Opportunity to decouple, to the great extent, hardware and software development and testing from each other, greatly improves the flexibility of the work, but more importantly speed-up the project significantly.
Rather than designing and building scaled down versions of the converter, models of the hardware elements (switching cells, capacitors, inductors, breakers) were developed and deployed on the selected RT-HIL system. Even though this looks like simple tasks, there are various challenges associated with tricking the real controllers to execute control algorithms against the model of the hardware, rather than the hardware itself. Provided that controllers are existing, as it was in our case, interfacing them to the RT-HIL system requires development of custom interfaces and level conditioning of various analogue and digital signals.
Modelling of power electronic circuits for the RT simulation has its own difficulties, due to discrete (switched) nature of converter operation leading to a large number of possible switching states that impact evolution of the relevant state variables (e.g. currents of inductors and voltages of capacitors). While the active semiconductor devices are commutated by the external signals from the controllers, passive semiconductor devices need to be commutated by circuit conditions detected by the RT-HIL models. All this needs to be done in a relatively short time periods, falling in the range of several microseconds. Thus, fidelity of models and update rates play an important role during selection of the RT-HIL platform.
Developed system is shown in Fig.3. Industrial controllers are used for the implementation of the control algorithms, connected through custom made interfaces to 7 RT-Boxes. One RT-Box hosts the model of the application the converter is serving, while each of the remaining 6 RT-Boxes host the model of an MMC branch comprising up to 8 submodules. In other words, the MMC with up to 48 submodules can be successfully simulated.
Nowadays, there are many vendors offering tailored RT-HIL systems for power electronic applications. These systems have come long way forward in terms of their performances and have become trusted and essential tool, providing important design support during research and development. They are increasingly replacing, or have replaced already, analogue based simulators in the industry. Ease of circuit modifications, change of test scenarios, simple variation of parameters, comfort of digital world, as only some of the advantages of digital RT-HIL systems.
Example presented in this article is only one way of how RT-HIL system can be used, and various other options are available, depending on what is the problem at hand. Testing power electronics control software is becoming easier and significant time and money can be saved, during the development, with the selection of appropriate tools.