Hardware and software for power system and market studies using AI and bigdata techniques

点击翻译按钮获取中文摘要

Large interconnected electric power systems are strategic for the economic and social well-being of Canadians, since society depends more than ever on electricity, due to the fact that most of our energy demand, in particular thermal and transportation, are being electrified at a very rapid rate, driven by climate change and the associated integration of ever more present and cheaper Renewable Energy Sources (RES) such as wind and solar generation. These large and complex grids, which are considered the most complex systems invented by humanity, have traditionally been studied and researched through software simulations, rather than hardware-based studies. In this context, our research group at Waterloo has focused on power systems research that is primarily based on computer modeling and simulation of mathematical models of large and small power grids with RES for stability, control, operation, economics, planning, and electricity market studies. For over the past two decades, our research group has developed some of the most relevant mathematical models for computer simulations and studies in the power grid technical literature, as attested by multiple awards and recognitions, which has been widely referred to and used by researchers and utilities around the world. Our research has been accomplished with the help of 4 servers and various software packages, on which our group is fully dependent. All this hardware and software tools, which have been developed, purchased, maintained, and updated with NSERC, OCE, research contracts, and Waterloo funds, have served us well, allowing us to also establish collaborations with other colleagues at Waterloo and other Canadian universities, as well as with researchers from overseas institutions. However, these servers and power system modeling and simulation software packages are becoming outdated and hence need urgent upgrade. Furthermore, the software versions currently available in the market have many advanced features that our existing versions are lacking. In fact, our oldest server has become rather slow in performing the required processing tasks because of its age and overuse, which is affecting the quality and quantity of our research output. In view of the above major constraints, we are applying for the Research Tools and Instrument (RTI) funds to replace the aforementioned server with an up-to-date more advanced version, especially since we are starting to pursue research in artificial intelligence (AI), machine learning (ML), and big-data applications to power systems. This evolving new research direction being supported by upcoming grants and contracts would require a different type of server based on multiple Graphic Processing Units (GPUs), which are indispensable for executing AI-related software. GPUs are the best options for fast ML as data science model training involves calculations that are greatly enhanced by parallel computations.